<%BANNER%>

Impact of Medicare Non-Payment of Hospital-Acquired Infections on Patient Outcomes and Hospital Financial Performance

Permanent Link: http://ufdc.ufl.edu/UFE0042716/00001

Material Information

Title: Impact of Medicare Non-Payment of Hospital-Acquired Infections on Patient Outcomes and Hospital Financial Performance
Physical Description: 1 online resource (143 p.)
Language: english
Creator: PEASAH,SAMUEL KWAME
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: FLORIDA -- HOSPITAL -- ICD -- INFECTIONS -- MEDICARE -- NOSOCOMIAL
Health Services Research, Management, and Policy -- Dissertations, Academic -- UF
Genre: Health Services Research thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: IMPACT OF MEDICARE NON-PAYMENT OF HOSPITAL- ACQUIRED INFECTIONS ON PATIENT OUTCOMES AND HOSPITAL FINANCIAL PERFORMANCE By Samuel Kwame Peasah May 2011 Chair: Niccie Lee McKay Major: Health Services Research Medicare, as part of the value purchasing program, implemented the non-payment of hospital-acquired conditions on October 1, 2008. The initial policy targeted 8 conditions including 3 hospital-acquired infections, which normally result in higher reimbursement rates from Medicare and are expensive to treat. The goal of this dissertation was to evaluate the impact of this policy on patient outcomes and the financial performance of acute-care hospitals in Florida. The study was designed to compare the before and after probabilities that a patient with any of these conditions acquired it while in the hospital. The study used discharge data from the Florida Agency for Health Care Administration, the Florida Hospitals Uniform Reporting System financial data, the Area resource file, and the American Hospital Association dataset among others. The patient outcomes analysis used generalized hierarchical linear model due to the dichotomous dependent variable and the clustering at the hospital level. OLS and mixed models were used for the financial performance question. Results indicated a decrease in the probability of a patient getting hospital-acquired infections and/or hospital-acquired conditions post-policy, but the reduction in the probability for catheter-associated urinary tract infections alone was not statistically significant. There was no evidence that revenues or margins of hospitals decreased due to the policy but inpatient care expenses increased post-policy.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by SAMUEL KWAME PEASAH.
Thesis: Thesis (Ph.D.)--University of Florida, 2011.
Local: Adviser: McKay, Niccie L.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0042716:00001

Permanent Link: http://ufdc.ufl.edu/UFE0042716/00001

Material Information

Title: Impact of Medicare Non-Payment of Hospital-Acquired Infections on Patient Outcomes and Hospital Financial Performance
Physical Description: 1 online resource (143 p.)
Language: english
Creator: PEASAH,SAMUEL KWAME
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: FLORIDA -- HOSPITAL -- ICD -- INFECTIONS -- MEDICARE -- NOSOCOMIAL
Health Services Research, Management, and Policy -- Dissertations, Academic -- UF
Genre: Health Services Research thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: IMPACT OF MEDICARE NON-PAYMENT OF HOSPITAL- ACQUIRED INFECTIONS ON PATIENT OUTCOMES AND HOSPITAL FINANCIAL PERFORMANCE By Samuel Kwame Peasah May 2011 Chair: Niccie Lee McKay Major: Health Services Research Medicare, as part of the value purchasing program, implemented the non-payment of hospital-acquired conditions on October 1, 2008. The initial policy targeted 8 conditions including 3 hospital-acquired infections, which normally result in higher reimbursement rates from Medicare and are expensive to treat. The goal of this dissertation was to evaluate the impact of this policy on patient outcomes and the financial performance of acute-care hospitals in Florida. The study was designed to compare the before and after probabilities that a patient with any of these conditions acquired it while in the hospital. The study used discharge data from the Florida Agency for Health Care Administration, the Florida Hospitals Uniform Reporting System financial data, the Area resource file, and the American Hospital Association dataset among others. The patient outcomes analysis used generalized hierarchical linear model due to the dichotomous dependent variable and the clustering at the hospital level. OLS and mixed models were used for the financial performance question. Results indicated a decrease in the probability of a patient getting hospital-acquired infections and/or hospital-acquired conditions post-policy, but the reduction in the probability for catheter-associated urinary tract infections alone was not statistically significant. There was no evidence that revenues or margins of hospitals decreased due to the policy but inpatient care expenses increased post-policy.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by SAMUEL KWAME PEASAH.
Thesis: Thesis (Ph.D.)--University of Florida, 2011.
Local: Adviser: McKay, Niccie L.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0042716:00001


This item has the following downloads:


Full Text

PAGE 1

1 IMPAC T OF MEDICARE NON PAYMENT OF HOSPITAL ACQUIRED INFECTIONS ON PATIENT OUTCOMES AND HOSPITAL FINANCIAL P ERFORMANCE By SAMUEL KWAME PEASAH A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN P ARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011 Samuel Kwame Peasah

PAGE 3

3 To my wife, Mrs. Hannah Abrefi Peasah, whose sacrifice, encouragement, and love saw me through to the end, an d to the memory of my late mother, Madam Regina Kyerewaa Danquah, who started it all, may her soul rest in peace

PAGE 4

4 ACKNOWLEDGMENTS I am grateful to God for seeing me through this journey. May His name be praised forever. So many people contributed to thi s successful story. Firstly, I want to thank Dr. Niccie Lee McKay, my chair and advisor, for her guidan ce and dedication which were instrumental to the successful completion of my program. Secondly, I want to thank Dr. Jeffery Harman, a committee member, w hose advice led to the choice of study design and the interpretation of the results. Thirdly, I want to thank my other two committee members; Dr. Mona Al Amin and Dr. Robert Cook for their support throughout the dissertation. I am also grateful to the Agen cy for Health Care Administration who gave me the bulk of data I used for this dissertation, especially Ms Arlene Schwahn for her promptness and p rofessionalism. Lastly, I want to thank my family; Hannah my wife, Sam Jr my son, and Anabel my daughter, for putting up with my long hours behind the computer, and my other family members and friends for their encouragement and support.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 LIST OF TABLES ................................ ................................ ................................ ........................... 8 LIST OF FIGURES ................................ ................................ ................................ ......................... 9 ABSTRACT ................................ ................................ ................................ ................................ ... 11 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .................. 13 Background ................................ ................................ ................................ ............................. 13 Research Questions ................................ ................................ ................................ ................. 15 Rati onale for the Study ................................ ................................ ................................ ........... 15 Contribution of This Study ................................ ................................ ................................ ..... 17 2 LITERATURE REVIEW ................................ ................................ ................................ ....... 19 Infection Management before the 20th Century ................................ ................................ ..... 19 Infection Control Management since the 20th Century ................................ .......................... 21 Effects of Pa yment Reforms ................................ ................................ ................................ ... 24 Economic Stabilization Program ................................ ................................ ............................ 24 Prospective Payment for Hospital Services ................................ ................................ ............ 25 Prospective Payment for Other Services ................................ ................................ ................ 28 Balanced Budget Act of 1997 ................................ ................................ ................................ 29 Pay For Performance ................................ ................................ ................................ .............. 32 Summary: Payment Reforms ................................ ................................ ................................ .. 35 3 CONCEPTUAL FRAMEWORK ................................ ................................ ........................... 37 Framework Linking Financial Incentives to Patient Outcomes ................................ ............. 38 Donabedian Structure Process Outcome Model ................................ ............................ 38 Principal Agent Theory ................................ ................................ ................................ ... 41 Production of Quality ................................ ................................ ................................ ...... 43 Effectiveness of Financial Incentives ................................ ................................ .............. 44 Fr amework Linking Financial Incentives to Financial Performance ................................ ..... 4 7 Neoclassical Theory of the Firm ................................ ................................ ..................... 48 Factors Affecting Hospital F inancial Performance ................................ ......................... 49 The Profitability Model ................................ ................................ ................................ ... 50 Hypothesis ................................ ................................ ................................ .............................. 52 4 METHODOLOGY ................................ ................................ ................................ ................. 55

PAGE 6

6 Study Design ................................ ................................ ................................ ........................... 55 Overview of All Data Sources ................................ ................................ ................................ 55 Florida Hospital Uniform Reporting System ................................ ................................ .. 56 Area Resource File ................................ ................................ ................................ .......... 57 Hospital Compare ................................ ................................ ................................ ............ 57 America Hospital Association Dataset ................................ ................................ ............ 57 Hospital Blue Book ................................ ................................ ................................ ......... 57 Empirical Specification ................................ ................................ ................................ .......... 57 Hospital Acquired Infections ................................ ................................ ........................... 58 Financial Incentive ................................ ................................ ................................ .......... 58 Financial Performance ................................ ................................ ................................ ..... 59 Variable Descriptions ................................ ................................ ................................ ............. 60 Dependent Variables ................................ ................................ ................................ ....... 60 Dependent vari ables for the patient outcomes question ................................ ........... 60 Dependent variables for the financial performance question ................................ ... 61 Independent Variables ................................ ................................ ................................ ..... 62 For the patient outcome question: ................................ ................................ ............ 62 For the financial performance question ................................ ................................ .... 62 Covariates ................................ ................................ ................................ ........................ 62 Structure measures ................................ ................................ ................................ ... 62 Process measures ................................ ................................ ................................ ...... 63 Other Covariate Measures ................................ ................................ ............................... 64 Covariates for the Financial Performance Question ................................ ........................ 64 Organizational measures ................................ ................................ .......................... 64 Managerial measure ................................ ................................ ................................ 64 Patient mix measures ................................ ................................ ............................... 65 Market measures ................................ ................................ ................................ ...... 65 Econometric Methods ................................ ................................ ................................ ............. 65 Patient Outcome Model ................................ ................................ ................................ ... 66 Financial Performance Mode l ................................ ................................ ......................... 66 5 RESULTS ................................ ................................ ................................ ............................... 71 Part 1: Patient Outcome Analysis ................................ ................................ ........................... 71 Descriptiv e Statistics ................................ ................................ ................................ ....... 71 ANOVA Analysis ................................ ................................ ................................ ............ 72 Multivariate Analysis ................................ ................................ ................................ ...... 75 Indep endent Variables ................................ ................................ ................................ ............ 76 Other Covariates the CAUTI Model ................................ ................................ ....................... 76 HAI Model ................................ ................................ ................................ ....................... 77 HAC Model ................................ ................................ ................................ ..................... 77 Independent Variables ................................ ................................ ................................ ..... 79 Patient characteristics ................................ ................................ ............................... 80 Hospital characteristics ................................ ................................ ............................ 82 Intercept ................................ ................................ ................................ .................... 82 Sensitivity Analysis ................................ ................................ ................................ ......... 83 Part 2: Financial Performance of Acute Care Hospitals ................................ ......................... 85

PAGE 7

7 Descriptive Statistics ................................ ................................ ................................ ....... 85 ANOVA Analysis ................................ ................................ ................................ ............ 86 Direct inpatient care revenues (LNDIR) ................................ ................................ .. 87 Direct inpatient care expenses PCEXP ................................ ................................ .... 88 Ne t patient care revenues (LREV) ................................ ................................ ........... 88 Net patient care expenses (LCOST) ................................ ................................ ......... 88 Operating profit (OPMARGIN) ................................ ................................ ............... 89 Profit margin (MARG) ................................ ................................ ............................. 89 Multivariate Analysis ................................ ................................ ................................ ...... 90 Covariates of operating margin and operat ing profit ................................ ............... 91 Covariates of net revenue and inpatient revenue ................................ ...................... 91 Covariates of inpatient expenses and total operating expenses ................................ 92 Intercept ................................ ................................ ................................ .................... 92 Sensitivity Analysis ................................ ................................ ................................ ................ 93 6 DISCUSSION ................................ ................................ ................................ ....................... 127 Summary ................................ ................................ ................................ ............................... 127 Hypothesis: Patient Outcomes ................................ ................................ ....................... 127 Hypotheses: Financial Perfo rmance ................................ ................................ .............. 128 Other Interesting Findings ................................ ................................ ................................ .... 130 7 CONCLUSION ................................ ................................ ................................ ..................... 132 Limit ations of the Study ................................ ................................ ................................ ....... 133 Policy Implications ................................ ................................ ................................ ............... 134 Future Research ................................ ................................ ................................ .................... 135 Final Thoughts ................................ ................................ ................................ ...................... 135 LIST OF REFERENCES ................................ ................................ ................................ ............. 136 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ....... 143

PAGE 8

8 LIST OF TABLES Table p age 1 1 T he charges and prevalence of selected hospital acquired infections in 2007 .................. 18 4 1 Variables for the patient out comes question. ................................ ................................ ..... 68 4 2 Variables for the financial performance question. ................................ ............................. 69 5 1 Means of Level 1 Variables, pre S ................................ ................................ ................................ ............................ 95 5 2 Means of Level 2 Variables per year; 2007, 2008, 2009. ................................ .................. 96 5 3 Bivariate odds ratio of dependent variables by policy ................................ ....................... 97 5 4 GHLM Regression results of main effects of patient outcome variables ........................ 104 5 5 GHLM Regression results of full model for patient outcome variables .......................... 105 5 6 Odds ratio for full model ................................ ................................ ................................ .. 107 5 7 Odds ratio for full mod el of the categorized age and Medicare admissions model ......... 108 5 8 Means of financial question variables per year; 2007, 2008, 2009 ................................ 110 5 9 Results of main effects of RHAC (hospital acquired conditions not present on admission) on the financial performance of acute c are hospitals ................................ .... 122 5 10 Results for full model effects o f RHAC (hospital acquired conditions not present on admission) on financial performance of acute care hospitals ................................ .......... 123 5 11 Results for main effects model of HAC (hospital acquired conditions pre sent or not present on admission) on financial performance of acute care hospitals. ..................... 124 5 12 Results for main effects of RHAC (categorized into 4 groups) and MCARE (categorized into 4 groups ) on financial performance of acute care hospitals ................ 125 5 13 Results for full model effects of categorized RHAC and MCARE on financial performance of acute care hospitals ................................ ................................ ................ 126

PAGE 9

9 LIST OF FIGURES Figure page 3 1 Kunkel Structure Process Outcome Models ................................ ................................ ...... 53 3 2 The Financial Incentiv e Structure Process Outcome model ................................ ............. 54 5 1 CAUTI for MEDICARE Groups (1, 2 & 3) vs. POLICY (0, 1) ................................ ........ 97 5 2 CAUTI for MEDICARE Group 1 (<25%), 2 (25 50%), & 3 (>50%) vs. YEAR (2007 to 2009) ................................ ................................ ................................ .............................. 98 5 3 HAI for MEDICARE Groups 1, 2, &3 vs. POLICY (0,1) ................................ ................ 98 5 4 HAI for MEDICARE Groups 1 (<25%), 2 (25 50%), & 3 (>50%) vs. YEAR (2007 to 2009) ................................ ................................ ................................ .............................. 99 5 5 HAC for MEDICARE Groups 1 (<25%), 2 (25 50%), & 3 (>50%) vs. POLICY ............ 99 5 6 HAC for MEDICARE Group 1 (<25%), 2 (25 50%), & 3 (>50%) vs. YEAR (2007 to 2009) ................................ ................................ ................................ ............................ 100 5 7 CAUTI for MEDICARE Groups 1 ( <25%), 2 (25 50%), & 3 (>50%) vs TREND ....... 101 5 8 HAI for MEDICARE Groups 1 (<25%), 2 (25 50%), & 3 (>50%) vs. TREND ............ 102 5 9 HAC for MEDICARE Groups 1 (<25%), 2(25 50%), & 3 (>5 0%) vs. TREND ............ 103 5 10 Partial results from categorized model of HAC ................................ ............................... 111 5 11 Partial results of inpatient revenues ................................ ................................ ................. 112 5 12 Log Inpatient Revenue vs. Medicar e year Infection rate ................................ ................. 112 5 13 Partial results of inpatient expenses ................................ ................................ ................. 113 5 14 Log Pat ient care expenses vs. Medicare year Infection rate ................................ ............ 114 5 15 Partial results of net patient revenue ................................ ................................ ................ 115 5 16 Net Revenue vs Medicare Infection rate ................................ ................................ .......... 116 5 17 Partial results of net cost ................................ ................................ ................................ .. 11 7 5 18 LCOST vs. M edicare year RHAC ................................ ................................ ................... 118 5 19 Partial results of operating margins ................................ ................................ ................. 119

PAGE 10

10 5 20 Operating profit vs. Medicare Year Infection rate ................................ ........................... 119 5 21 Partial results of profit margin ................................ ................................ ......................... 120 5 22 Profit Margin vs. Medicare year Infection rate ................................ ................................ 121

PAGE 11

11 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy IMPACT OF MEDICARE N ON PAYME NT OF HOSPITAL ACQUIRED INFECTIONS ON PATIENT OUTCOMES AND HOSPITAL FINANCIAL P ERFORMANCE By Samuel Kwame Peasah May 2011 Chair: Niccie Lee McKay Major: Health Services Research Medicare, as part of the value purchasing program, implemented the non payment of hospital acquired conditions on October 1, 2008. The initial policy targeted 8 conditions including 3 hospital acquired infections, which normally result in higher reimbursement rates from Medicare and are expensive to treat. The goal of this dissertation was to evaluate the impact o f this policy on patient outcomes and the financial performance of acute care hospitals in Florida. The study was designed to compare the before and after probabilities that a patient with any of these conditions acquired it while in the hospital. The stud y used discharge data from the Florida Agency for Health Care Administration, the Florida Hospitals Uniform Reporting System financial data, the Area resource file, and the American Hospital Association dataset among others. The patient outcomes analysis used generalized hierarchical linear model due to the dichotomous dependent variable and the clustering at the hospital level. OLS and mixed models were used for the financial performance question. Results indicated a decrease in the probability of a patie nt getting hospital acquired infections and/or hospital acquired conditions post policy, but the reduction in the probability for catheter associated urinary tract infections alone was not statistically significant. There was no

PAGE 12

12 evidence that revenues or m argins of hospitals decreased due to the policy but inpatient care expenses increased post policy.

PAGE 13

13 CHAPTER 1 INTRODUCTION Background Access to health care, cost of health care, and quality of health care are the three key areas that inform debate about r eform of the U S health care system. After the elderly, the poor, and those with various forms of disability were given access to health care through the passage of legislation authorizing the Medicare and Medicaid programs in the mid 1960s, the focus shi fted to payment structure in efforts to contain escalating costs. Recently quality of care has received increased attention following reports of disparities th e attention of lawmakers and the public to the seriousness of medical errors. For example, patients who acquire an infection in the hospital were found to be 6 times more likely to die, than patients who acquire similar infections outsi de the hospital (IHM 2009). Overall, about 100,000 deaths occurring in our hospitals annually are preventable according to the IOM report. Hospital acquired infections are especially common, with approximately 1.7 million hospital acquired infections occurring nationally in 2 002 and about 98,987 deaths resulti ng from the infections (Klevens et al. 2002). Medical errors also have significant consequences for health care expenditures. For example, the annual direct hospital cost of treating health care associated infections in the United States was estimated to fall in the range of $28.4 to $33.8 bill ion (Scott 200 2 ). Another study Medicare hospital payments, ranging from an average of add itional $700 per case of decubitus ulcers to $9,000 per case to treat post operative sepsis (CMS 2006).

PAGE 14

14 These findings led to calls to restructure our payment systems to promote quality of care. Congress responded to the public outcry with provisions in two pieces of legislation: the Medicare Modernization Act of 2003 and the Deficit Reduction Act of 2005. Portions of these Acts authorized the Secretary of Health and Human Services to incorporate some of the suggestions of the IOM to create a value purchasin g program for Medicare services. The Centers for Medicare and Medicaid Services (CMS), the federal agency responsible for implementation of these provisions, set up a plan for implementing the hospital value purchasing program. As part of the plan, CMS a nnounced that, beginning October 1, 2008, the agency would not reimburse hospitals for Medicare claims reflecting certain types of hospital acquired infections and hospitals would not be allowed to bill Medicare patients directly for treating such infectio ns. As defined by CMS, hospital acquired infections (nosocomial infections) are infections that were not present at the time of admission and that present after 48 hours of admission. As of F iscal Year 2009 (October 1, 2008 to September 30, 2009), CMS wil l not reimburse for catheter associated urinary tract infections (CAUTI), vascular catheter associated infections (VCAI), or mediastinitis, (a type of surgical site infection) after coronary bypass graft (CABG). Table 1 1 presents the number of cases and c harges per hospital stay for these three conditions in 2007. Non payment by Medicare for these types of nosocomial infections is a major reform in the way CMS reimburses hospitals for services to Medicare patients. Given that previous payment reforms were found to affect hospital behavior and performance, an important question is how this non payment policy impacts hospitals.

PAGE 15

15 Research Questions To address this important policy issue, this dissertation will examine the effect of the non payment policy on pat ient outcomes and hospital financial performance. The specific research questions to be addressed are: 1. What is the impact of the Medicare non payment policy on patient outcomes (the likelihood of patients acquiring the targeted hospital acquired infections CAUTI, SSI, and VCAI)? 2. What is the impact of the Medicare non payment policy on the financial performance of short term acute care hospitals? Rationale for the Study The use of financial incentives to influence behavior is becoming more common in the hea lthcare sector. Virkstis et al (2009) report that within the past five years privately sponsored pay for performance programs have increased from 40 to 150 programs. However, Mehrotra et al. (2009), Petersen et al. (2006), and Glickman et al. (2007), amo ng others, have commented on the inconclusiveness of the impact of pay for performance programs. Some of the reasons given include the paucity of research in this area and the diversity of financial incentives being used. For example, traditionally incenti ves have involved bonus payments, but CMS is shifting to a strategy that withholds payment for non performance. This change in strategy stems from the language of the Deficit Reduction Act of 2005 which aims at controlling both quality and cost. The Defic it Reduction Act of 2005 that relates to are (a) high cost or volume or both; (b) result in the assignment of a Diagnosis Related Group (DRG) that has a h igher payment when present as a secondary diagnosis; and (c) could reasonably have been prevented et al. 2009, p. 351). CMS has identified several of such conditions for consideration.

PAGE 16

16 For Fiscal Year 2009, eight conditions were identified for the non payment policy and CMS has advised that several other conditions are under consideration for the future including: extreme blood sugar derangement, clostridium difficile associated infections, and surgical site infections occurring after orthopedic an d bariatric surgeries (Catalano 2008). Evaluation of the 2009 non payment policy for hospital acquired conditions is important because the findings could be used to modify future policies as needed. To date only two studies, both simulations, have exami ned the effect of Medica payment policy. McNutt et al (2010), using discharges from all patients in 86 academic medical centers across the U S between July 2005 and June 2007, estimated the average amount of reimbursement loss per hospital due to the Medicare non payment policy. The study found that at least 4% of patients had at least one of the targeted eight hospital acquired conditions and estimated the average amount of reimbursement loss per hospital per year at approximately $1 million. Mc Nair et al (2009) based their estimation on a simulation model using the California Office of State Health Planning and Development 2006 Patient Discharge Dataset. The study estimated that the Medicare non payment policy would result in a reimbursement lo ss per hospital per year between $92,000 and $227,000, substantially lower than the estimate by McNutt et al (2010). The study by McNair et al (2009) drew reaction from CMS officials due to speculation in the article that the financial impact would be to o small to elicit any changes in hospital behavior. reduction 1494).

PAGE 17

17 Contribution of This Study The impact of PPS and BBA on patient out comes and hospital financial performance is well established. The use of payment withholds to influence behavior change is in its nascence. Did the financial pressure created by the Medicare non payment policy impact patient outcomes as intended? Was it st rong enough to impact hospital financial performance as an incentive to spur behavioral change? This dissertation will answer these questions and contribute to the literature on the use of financial performance (specifically payment withholders) to influen ce hospital behavior. No study to date has used discharge data from the post policy period to assess the impact of the non payment policy on patient outcomes and the financial performance of hospitals. This dissertation used a pre post study design, modele financial performance. This study contributes to our understanding of the impact of Medicare revenue reductions on targe ted patient outcomes and its impact on the financial performance of hospitals. This study also contributes to the evaluation of the Medicare non payment policy as the policy considers the possibility of expanding the number of conditions. In summary, the p ublic outcry over the IOM report on the prevalence and preventability of medical error led to the push by congress for payment reform that incorporates patient outcomes. Medicare non payment policy is one of the payment reform steps taken by CMS to challen ge hospitals to better patient outcomes. Although pay for performance is gaining traction in the health sector, its impact is still inconclusive. This study will contribute to the debate and inform policy modification as we continue to use financial incent ives to influence provider behavior.

PAGE 18

18 Table 1 1. The charges and prevalence of selected hospital acquired infections in 2007 Condition Number of cases Charge per hospital stay Catheter associated UTI 12,185 $44,043 Vascular catheter associated infection 29,536 $103,027 Surgical site infection mediastinitis after (CABG) 69 $299,237 From the Federal Register (2008 p. 23551)

PAGE 19

19 CHAPTER 2 LITERATURE REVIEW The Medicare nonpayment policy is predicated on empirical evidence that hospital acquired infections a re preventable, on the supposition that hospitals are in a position to improve patient outcomes if they engage in evidence based medical practices, and on the assumption that financial incentives will influence hospital behavior. This chapter will summariz e the advances made in the fight against infections up to the twentieth century, then focus on advances made since the twentieth century to control infections. The final portion of the chapter summarizes the research literature examining the impact of prev ious payment reforms on patient outcomes and hospital financial performance. Infection Management before the 20th Century Societal expectations regarding the competence and accountability of medical practitioners and the safety of patients date back at lea st to 1750 BC. Surgeons in Mesopotamia under the Code of Hammurabi were compensated for their services but their hands were amputated if their patients died under their care (Miller, Scott, and Lee 2005). Under this code, few surgeries were performed and surgery was considered a last resort. It was not until recently that scientists understood that most of the deaths from surgery were caused by infections. For example, during the last part of the 18 th century and the first part of the 19 th century, putrid wounds and numerous serious illnesses were thought to be caused by 1985) based on the fact that amputated patients recovered faster when treated at home than in the hospital and that when healthy individuals came into contac t with sick people in the hospitals they became more prone to illnesses. Hieronymus Fracastorius (1478 1553) was the first to postulate that infection was caused ct

PAGE 20

20 with infected persons, indirect contact with fomites, and air Scott, and Lee 2005 ). The discovery of the microscope by Von Leeuwenhoek in 1693 led to the discovery of the existence of microbes, and the link between microbes and infection. But the understanding of the importance of hygiene was slow. As late as 1863, Florence Nightingale was able to substantially reduce mortality rates at military hospitals by us ing hygienic procedures (Simkin 1997). The father of modern hospi tal hygiene is widely considered to be the Hungarian obstetrician, Ignaz Semmelweis (1818 1865). He had observed at the Vienna Hospital that mortality rates from deliveries done by physicians and medical students in the hospital were higher th an those done by midwives (Best and Neuhauser 2004). After his friend died after getting pricked with a needle while he was performing an autopsy on a woman who had died from puerperal sepsis, Semmelweis suspected that there might be a correlation between his death and of infection from the dead woman (Alexander 1985). Thereafter he required all physicians and medical students to wash their hands with an antiseptic after an autop sy before delivering babies. The puerperal mortality rate fell from 9.92% to 1.27% in 2 years with the introduction of hand washing with chloride of lime solution (Miller Scott, and Lee 2005 ). The next significant advancement in the fight against infectio n came from the discovery by Louis Pasteur (1822 1895) that living microscopic organisms could cause fermentation and putrefaction of meat. This knowledge was the genesis of what later became popularly known as the germ theory. The germ theory led Joseph L ister in 1867 to establish the link between microbes and wound suppuration. He suspected that these organisms might be the cause of the

PAGE 21

21 mortality due to amputat ion from 45% to 15% simply by applying carbonic acid to compound fractures (Miller Scott, and Lee 2005 ). Robert Koch furthered our understanding of infections by demonstrating that different germs caused different diseases and also discovered the usefulnes s of steam sterilization of surgical instruments in 1881. Several other scientists contributed to some of the infection control practices used today. For example, Gustav Neuber introduced the use of sterile gowns and caps in 1883, Mikulicz introduced the surgical face mask in 1897, and in 1890 William Halsted made the use of rubber gloves popular when he commissioned the Goodyear rubber company to fashion gloves for his nurses to use during disinfection (Miller Scott, and Lee 2005 ). Infection Control Mana gement since the 20th Century In the 20 th century, Alexander Fleming made the most significant contribution to the fight against infection in 1928 with the accidental discovery of the antibiotic, penicillin. This initiated the use of antibiotics as prophyl axis against infection during surgical operations and as treatment for various forms of infections. Today, surgeons routinely give an antibiotic as prophylaxis one hour before surgical procedures. A national outbreak of hospital based staphylococcal infec tion in 1957 and 1958 brought public health officials into the center of the study and control of hosp ital acquired infections (Burke 2004). Almost 20 years later, in 1976 the Joint Commission on Accreditation of Healthcare Organizations (JCAHO) included p revention of hospital acquired infections as one of its patient safety goals (Anderson et al. 2007). This resulted in the active involvement of hospital administrators and the development of infection control units in hospitals. In 1985 the Centers for Di sease Control (CDC) reported the findings of a 10 year study (1974 to 1984) called the Study on the Efficacy of Nosocomial Infection Control (SENIC) (Haley et al. 1980). The goal of the study was to assess the various strategies hospitals use in

PAGE 22

22 controllin g hospital acquired infections and to evaluate the effectiveness of these approaches. The final report attributed effectiveness of infection control to the presence of these four factors: the existence of an active surveillance program (a program of collec ting, analyzing, and dissemination of findings); control measures (a program that institutes measures to implement the findings from the surveillance with feedbacks); an infection control practitioner (one controller per 250 occupied beds); and a trained p hysician in epidemiology. Overall, infection rates were 32% lower for hospitals that had these four factors in place. Because duration of catheterization is widely recognized as an important factor in developing infection, JACHO recommends limited use and monitoring as a control measure. A study by Apisarnthanarak et al (2007), found that in hospitals where nurses are daily alerted to remind physicians to follow up on urinary catheters of patients, infections per 1000 catheter days fell 21.5% to 5.2%. Ano ther preventive strategy found to control catheter induced infection is the use of coated catheters. A systematic review of randomized and quasi randomized control studies confirmed the effectiveness of coated urinary catheters, and concluded that nitrofur azone coated catheters were superior to silver alloy coated catheters in preventing infections (Johnson, Kuskowski, and Wilt 2006). The fight to control hospital acquired infections is a global phenomenon and has captured the attention of several national health care systems and the World Health Organization. A World Health Organization report in 2002 defined hospital acquired infections as infections acquired during hospital care which are not present or incubating at admission. Infections occurring more than 48 hours after admission are usually considered hospital acquired or nosocomial. The report identified the following factors as leading to nosocomial infections; the microbial agent, patient susceptibility, environmental factors, and bacterial resista nce. The report

PAGE 23

23 concludes that nosocomial infections can lead to functional disability, emotional stress, mortality, economic costs due to increased length of stay, and additional use of resources including drugs, laboratory/other diagnostic testing. Much progress has been made over the years to either control or prevent these infections; the most has been achieved through environmental preventi on and use of antibiotics (WHO 2002) The effectiveness of antibiotics in controlling infection unfortunately led to overuse resulting in antibiotic resistance. The most common of these resistant bacteria in the hospital is methicillin resistant staphylococci aureus (MRSA). Although hospital acquired MRSA is a growing problem, the most frequently encountered hospital acquired infection is catheter associated urinary tract infections, accounting for about 35% of all nosocomial infections, followed by surgical site infections which account for about 20% of all infections (these two infections are the subject of this dis sertation). However, these two nosocomial infections are not as expensive to treat as blood stream infections and pneumonia, which account for about 15% each of a ll nosocomial infections (Burke 2004). A recent infection control program that has received a great deal of attention of congress is the Michigan Keystone Project. Physicians and researchers collaborated with the Agency of Healthcare Research and Quality (AHRQ) to conduct a cohort study in 108 intensive care units mainly in Mi chigan. The study us ed evidence based intervention recommended by the CDC and which the authors identified as to have the greatest effect on the infection in question catheter barrier pre caution during the insertion of central venous catheters, cleaning the skin with chlorhexidine, avoid the femoral site if possible, and removing unnecessary cath (Pronovost et al 2006, p. 2726). The results indicate a reduction in catheter related blood stream

PAGE 24

24 infection as early as 3 months after implementation. The results were sustained even 18 months post implementation. Effects of Payment Reforms Medicare, an entitlement program established by the Medicare Act of 1965, has had an enormous impac t on the delivery and funding of medical care in this country. When it was established in 1965, the focus was extension of medical coverage to the elderly. In order to get the medical community to accept and work with the program, provisions were incorpora ted to protect the financial security of providers. Consequently, hospitals were paid based on reasonable costs and physicians were reimbursed based on customary charges. In the first year of Medicare, physician fees more than doubled, while hospital charg es increased by 21.9% (Kulesher 2006). This pattern of escalating Medicare costs has continued ever since. To deal with increasing expenditures, various Medicare payment reforms have been implemented. This section summarizes the empirical evidence about th e impact of these payment reforms on patient safety and hospital financial performance. Economic Stabilization Program Establishment of the Medicare and Medicaid programs did more than increase access to seniors and low income beneficiaries, it also acted as an assurance to providers of payment for their services. Subsequently, inflation in the health sector increased faster than the rest of the economy and this was attributed mainly to increased prices, wages, and technolo gy in the medical sector (Gold an d Chu 1993). The Economic Stabilization Act (ESP) of 1970 was passed to control prices, wages, and rents. ESP rolled out in phases, the first being a 90 day freeze on wages. A limit of 5.5% on wage increases and other price control measures were aimed at c utting inflation in the whole economy in half.

PAGE 25

25 Gold and Chu (1993) reviewed the literature on the effects of selected cost containment efforts between 1971 and 1993. They report that hospital profit margins fell by about 6% following ESP, but the decline in profits was attributed to the mandated wage control but not necessarily operational efficiency. Although ESP did reduce inflation, it was allowed to expire in 1974, and subsequently inflation rose back to pre ESP levels with the hospital sector having t he greatest increase in consumer price index. Prospective Payment for Hospital Services Hospital care expenditures increased from $24.7 billion dollars in 1972 to $30.6 billion in 1982 (Kulesher 2006). In 1982, Medicare implemented the prospective payment system (PPS). The objective of PPS was to curb cost by moving away from the cost based (retrospective) payment system to a prospective payment system in which providers are paid based on the type of services provided. The diagnosis related groups (DRGs) we re designed to group together patients with similar medical conditions expected to use similar amounts of resources. Because hospitals received a fixed amount per discharge, they were at risk of losses if costs per discharge exceeded the amount of reimburs ement received. The literature on the impact of PPS examined the effect of the new payment system on a variety of factors, including hospital performance and patient outcomes. Feder Hadley, and Zuckerman l (1989) used 1982 to 1984 national data to examine the initial reaction of hospitals after PPS implementation. The study found that hospitals were still able to make profits because they reduced operational costs. Another study reviewed the literature covering five years post PPS and concluded that hospita ls reduced their operational cost mainly through reduced length s of stay per discharge (Chulis 1991). Fisher (1992) used Medicare cost reports and American Hospital Association survey data from 1985 (two years after PPS) to 1990 (seven years after PPS) to categorize hospitals into

PAGE 26

26 three groups: those that were profitable in 1985 and remained profitable in 1990, those that were profitable in 1985 but were not profitable in 1990, and those that were not profitable in 1985 but became profitable in 1990. The st udy found that the proportion of hospitals with Medicare PPS inpatient profits dropped steadily over the period from 84.5% to 40.7%. Hospitals that made Medicare PPS inpatient profits also had higher overall facility profits than hospitals that made Medica re PPS inpatient losses and had lower Medicare PPS inpatient expenses as compared to the hospitals with Medicare inpatient losses. Another review found overwhelming evidence that in the first year of PPS, hospitals made windfall profits, which were attribu ted to increases in the case mix index, reduced length of sta y, and cost reductions (Coulam and Gaumer, 1992). After the second year of PPS, however, hospital profits started declining and by the sixth year of PPS, 57% of hospitals had negative margins on Medicare inpatients (wh ich were similar to margins pre 1980). In addition, the number of hospital closures four years post PPS were more than double the hospital closures four years prior to PPS, with most of the post PPS hospital closures being rural hosp itals with less than 100 beds. Steinwald and Dummit (1989) reviewed data from the Prospective Payment Assessment Commission (ProPAC) about the impact of PPS on hospital case mix and found that the case mix index (CMI) increased by 6% after the first year of PPS. Hospitals are re imbursed according to diagnosis related group (DRG) which is based on the primary and secondary diagnosis and procedures (including subsequent complications and co morbidities) and patient age, sex, and discharge status. A survey co nducted by ProPAC of hospital medical personnel found that after the implementation of PPS, personnel paid more attention to completing and coding of medical

PAGE 27

27 records. Overall, from 1981 to 1988, cumulative payments due to case mix change s increased over 2 0% (Steinwald and Dummit 1989). A systematic review of the literature on impacts of PPS on hospital performance by Davis and Rhodes (1988) concluded that, although hospital utilization and average length of stay declined, quality of care did not decline as had been anticipated. However, there was a significant shift of procedures and patients from the inpatient setting to less costly outpatient settings. Khan et al (1990) found an improvement in quality after PPS, using national representative sample of Medicare patients admitted pre PPS (1981 82) and post PPS (1985 86). The study used 14,012 Medicare patients hospitalized within the given periods with one of five different diseases to assess the impact of PPS on in hospital mortality and length of stay, and found a 24% reduction in average length of stay and a declin e from 16.1% to 12.6% in in hospital mortality rates. Cutler (1995) on the other hand, found a negative association between PPS and patient outcomes. This was a longitudinal study of over 30, 000 Medicare patients in Massachusetts between 1981 and 1988. After computing price declines in hospitals due to PPS, the study concluded that hospitals with greater price declines also had a greater share of deaths either in the hospital or shortly after patients were discharged, although there were no differences in mortality one year post discharge. The study found that there was an increasing trend in readmissions post PPS. A study by Shen (2003) examined the relationship between financial pressure and quality of care, with financial pressure defined as a policy or an environment that puts a financial burden on hospitals. The study specifically considered both pressure from reduced reimbursement to

PAGE 28

28 hospitals due to PPS and also the reduced reimbursement to hospitals by managed care organizations from increased penetration in a given market. Mortality after acute myocardial infarction (AMI) treatment was used as a proxy for quality of care in the study. The study examined AMI data between 1985 and 1994 in short term non federal hospitals in the US. The results indicated that there was a negative association between financial pressure and short term mortality (30 day mortality rates), but differences in mortality rates after one year were not statistically s ignificant. In summary, research on the impact of the DRG payment system found that, in spite of increases in the Medicare case mix index and reductions in operating costs, hospital profits declined after the first year of the new payment system. Moreover, hospital closures increased after the introduction of PPS. In general, the consensus of the literature is that hospital admissions and average length of stay declined and that the new payment system resulted in a shift of patients from inpatient to outpat ient settings. Regarding effects on the quality of care, however, there is less agreement, with some studies finding unchanged or improved quality after PPS and others concluding that quality declined after PPS. Prospective Payment for O ther Services Imple mentation of PPS started with general hospitals because they constitute the bulk of Medicare expenditures. Eventually, however, PPS was extended to other healthcare organizations, including skilled nursing facilities (SNF), home health organizations, and r ehabilitation hospitals. This section briefly summarizes the impact of PPS on these facilities to augment understanding of the impact of policy on financial and quality performances. Inpatient rehabilitation hospitals continued under the cost based reimbur sement system until 2002 when they were switched to PPS. Thompson and McCue (2 010) examined the pre and post PPS periods for these hospitals to understand the impact of the change on their

PAGE 29

29 behavior and financial performance. The study compared 96 for prof it and 44 not for profit freestanding inpatient rehabilitation hospitals before and after 2002, when PPS was implemented. Using a pair wise comparison test they found that both types of ownership decreased their length of stay, increased discharges, and we re more profitable post PPS. For profit hospitals had higher percentage of Medicare discharges and a higher decrease in operating expenses than the not for profit hospitals (Thompson and McCue 2010). PPS was implemented in 2000 for home health agencies a fter a brief period of an interim payment system authorized by the BBA of 1997. Under the PPS, Medicare pays an agency a set amount per each 60 day episode of care based on one of the 80 different home health resource groups (HHRGs), which are similar to h ospitals DRGs. Collins et al (2007) undertook a retrospective case study to examine the impact of PPS implementation on the number of visits and physical performances of Medicare patients with total knee replacements. The study hypothesized that these age ncies have the financial incentive to reduce number of visits to these patients. They found that the number of visits did fall and the patients had less improvement in knee extension range of motion post PPS (Collins et al. 2007). Konetzka et al (2004) in vestigated the effects of PPS on skilled nursing facilities. They used staffing ratios and regulatory deficiencies as proxies for quality of care and examined the impact of PPS on associated rate changes. Data came from Medicare cost reports and the online survey certification and reporting (OSCAR) system for 1996 to 2000 for a before and after panel data study design. The study found that there was a negative correlation between implementation of PPS and quality of care in these nursing homes. Balanced Bud get Act of 1997 In the mid 1990s, there was renewed concern that further reductions in Medicare spending were necessary to forestall the Medicare trust fund from going bankrupt. Congress responded

PAGE 30

30 with the passage of the BBA of 1997 an act that was project ed to reduce Medicare payments to providers by $119 billion from 1998 to 2002, including a $2.3 billion reduction in graduate medi cal education funding (Phillips et al. 2004). In analyzing hospital cost reports for 1996, 1998, and 1999, Phillips et al. (20 04) reported that the passage of BBA had an enormous impact especially on teaching hospitals with 35% of teaching hospitals having negative operating margins in 1996 and a 24% decrease in graduate medical education margins between 1996 and 1999. The Univer sity of Pennsylvania Medical Center reported a $198 million deficit in 1999 and other academic medical centers reported losses exceeding $50 million in that same year. The study quoted a congressional budget office (CBO) report stating that BBA had oversho t Medicare spending reductions b y $88 billion dollars (Phillips et al. 2004). Bazzoli et al (2005) compared the short term effects of the BBA 1997 with that of PPS since hospitals faced financial pressures under both policies. Using data from 1996 to 1999 the study found that hospitals behaved in similar ways after both payment reforms: they expanded outpatient services to supplement inpatient revenues, made efforts to contain Medicare inpatient cost growth, and contained hospital staff growth. The major d ifference was that after implementation of BBA, hospital profit margins declined and there was an increase in inpatient usage. This was in contrast with the general consensus in the literature for post PPS where inpatient services were shifted to outpatien t settings resulting in a declining inpatient usage but increased margins in the initial years. Younis (2006) also examined hospital profitability post BBA by using return on assets as a measure of profitability. He did a pre post BBA study using 1996 and 1999 Medicare cost reports. This study also found an overall decline in hospital profitability post BBA 1997.

PAGE 31

31 Perhaps Gail Wilensky, the chair (1997 to 2001) of the Medicare Payment Advisory Commission (MedPAC ) summarized the financial impact of BBA 1997 best during her testimony before the subcommittees on health & environment and committee on commerce at the US House of Representat ives on July 19, 2000 (Wilenski 2000). In her report she testified that Medicare margins to inpatient, outpatient, skilled nu rsing, home health, inpatient rehabilitation and psychiatric facilities dropped aggregately from 9.8% in 1997 to 6.5% in 1998. Medicare margin for acute inpatient services alone fell from 17% to a historic low of 14.4% in 1998. Studies conducted using the Medicare cost reports indicated that the aggregate total in, which was less than half what it was in 1997. Hospitals with negative inpatient margins also rose from 23% in 1997 to 29% in 199 8. Apart from the financial impact on hospitals, empirical evidence shows that BBA affected quality of care as well. Lindrooth et al (1993) used a fiscal pressure index as a proxy for the financial pressure on hospitals due to reduced reimbursement and nu rse per day ratio as a quality measure. Using data from the Area Resource File and the America Hospital Association survey for the years 1996 to 2001 for all urban short term general hospitals, the study found that registered nurse (RN) and licensed practi cal nurse (LPN) staffing levels per adjusted patient days declined post BBA. Hospitals that were most susceptible to lower reimbursements had the highest decline in staffing levels. Clement et al (2007) were interested in the impact of patient payer on ho spital patient safety. The study used four patient safety indicators (PSI), as defined by the Agency of health care quality and research, to assess the impact of changes in reimbursement patterns over time. Using discharge data from 11 States, the study c oncluded that performance on these measures

PAGE 32

32 worsened between 1995 and 2000 for both private and Medicare patients. The impact of BBA, however, was less consistent; although there was a significant decline in some PSIs, one PSI did not show a statistically significant change. Using 1997 to 2001 discharge, financial, and death certificate data from Pennsylvania hospitals, Seshamani et al (2006) assessed the impact of BBA on postoperative mortality. The study classified hospitals into three groups: low (less than 1.8%), high (at least 3.5%) and middle, depending on the level of net revenue reduction since BBA implementation and examined the 30 day mortality rates from 8 postoperative complications. Findings indicated that hospitals with high BBA impact had sig nificantly worse mortality rates compared to those with low BBA impact. In sum, the BBA 1997 had a more severe impact on hospitals than PPS. Most researchers did not find significant changes in patient outcomes after PPS, but patient outcomes were worse af ter BBA. Hospitals lost significant revenues leading to declining overall margins under BBA and margins did not rebound like they did under PPS. Consequently, Congress responded in 1999 with payment adjustments that led to increased payments to hospitals ( BBA1999). Pay F or Performance Unlike PPS and the BBA of 1997, which were responses to escalating Medicare cost and the need to curb the growth, pay for performance programs seek t o reward adherence to evidence based practices. Following the Institute of M Congress, through the Medicare Modernization Act (MMA) of 2003 and the Deficit Reduction Act (DRA) of 2005, auth orized the creati on of a value based payment system that would provide incentives for improving patient safety.

PAGE 33

33 The Centers for Medicare and Medicaid Services (CMS), the agency responsible for implementing the provisions of the law, created a hospital value based purchasin g program. As part of the program CMS initiated demonstration projects to examine initial evidence regarding the impact of financial incentives on adherence to evidence based guidelines, which involve process measures that have been demonstrated to affect patient outcomes (for example giving aspirin at arrival and at discharge following a heart attack). An evaluation of the Premier Hospital Quality Incentive Demonstration (one of the demonstrations CMS undertook in partnership with the private sector) showe d that after the first three years of the project, there were general improvements in average composite quality scores in all five clinical areas from baseline: 87% to 96.1% for patients with AMI, 84.8% to 97.4% for patients with CABG, 64.5% to 88.7% for p atients with Heart Failure, 69.3% to 90.5% for patients with Pneumonia, and 84.6% to 96.9% for patients with H ip and Knee replacement (CMS 2008). Khan et al (2006) used data from the Premier demonstration project to examine the distribution of hospitals based on composite scores for 3 conditions (heart attacks, heart failures, and pneumonia). The study found significant variation among hospitals for a given condition. For example, rural hospitals performed poorly in heart attacks and heart failures as com pared to urban hospitals. However, rural hospitals outperformed urban hospitals in pneumonia. Major teaching hospitals, urban, and tax exempt hospitals were among the top performers while non teaching, rural, public, and investor owned hospitals were more likely to be among the bottom 20% of heart attack and heart failure composite scores. Major teaching hospitals (about 38% of them) were in the bottom 20% for pneumonia. The overall conclusion was that the financial impact of the demonstration project was m arginal, but could be significant depending on the bonus scheme used. Medicare Payment Advisory Commission ( MedPAC ) recommends setting

PAGE 34

34 up a fund from a 1% reduction in reimbursement to all hospitals and use it to reward hospitals who attain specified goal s instead of rewarding the top 20% achievers. The authors found that using MedPAC formula would result in a more positive impact on the financial performance of hospitals than the one from the CMS demonstration project (Khan et al. 2006). Public reporting is one of the strategies adopted by CMS in implementing the hospital value purchasing program described earlier. Public reporting is also one of the tools adopted by several states to promote quality of care in their hospitals. Lind e nauer et al (2007) co mpared hospitals in states with public reporting who were also part of the CMS demonstration project (payment for performance group) with hospitals in states with public reporting but who were not part of the CMS demonstration project (control group). Th e impact of pay for performance was measured by changes in adherence to 10 individual and 4 composite process measures over a two year period, and the study divided hospital into quintiles based on their process measure adherence score at the start of the p roject. Study results indicated that pay for performance hospitals in the lower quintile at baseline had increased their scores by an average of 16.1% at the end of the two year period and the pay for performance hospitals in the highest quintile increased their baseline values by only 1.9% as compared to the control group (Lind en auer et al. 2007). Grossbart (2006) found a similar trend when comparing hospitals in the demonstration project with control groups that is a marked improvement in the quality of c linical process delivery for demonstration hospitals. Most of the stories reporting success with pay for performance programs have focused on adherence to process measures, whereas results regarding the impact of pay for performance on outcome measures hav e been mixed. Glickman et al. (2007) studied the impact of pay for performance using data for hospitals who either participated in the Premier Demonstration

PAGE 35

35 Initiative or the CRUSADE Initiative in (1996). CRUSADE is the Can Rapid Risk Stratification of Uns table Angina Patients Suppress Outcomes with Early Implementation of the American College of Cardiology/American Heart Association guidelines national quality improvement initiative. CRUSADE hospitals participating in the Medicare demonstration Initiative were compared to other CRUSADE hospitals not in the Medicare demonstration Initiative (controls) to evaluate the impact of pay for performance on six composite acute myocardial process measures, in hospital mortality due to AMI, and unintended consequences The study did not find any significant differences between the two groups for either process measures or in hospital mortality, nor was there any evidence of negative unintended consequences due to the incentive (Glickman et al. 2007). A review of multip le studies conducted by Petersen et al (2006) concluded that, five out of six studies on physician level incentives and seven out of nine provider level financial incentives showed partial or positive effects on quality. On the other hand, Mehrotra et al (2009) concluded from a systematic review of the literature that very few empirical studies exist for drawing any convincing conclusions about the impact of pay for performance on quality of care. Summary: Payment Reforms In evaluating the empirical fin dings regarding the impact of payment reforms and financial incentives on patient outcomes and hospital financial performance, the first policies evaluated: ESP, PPS, and BBA, were designed to curb the rising cost of health care and thus either mandated pr ice and wage controls (ESP) or reduced payment to health care facilities through direct cuts (BBA) or the switch from cost based to prospective payment (PPS). Hospitals responded to these policies by changing business practices to forestall the impact of t he reduced revenues. The strategies they adopted included reducing the length of stay, shifting some inpatient procedures to the outpatient setting, staff reductions, and increased attention to patient

PAGE 36

36 medical records and coding. These strategies resulted in an initial increase in profitability which ebbed with time under PPS. Some hospitals closed and others merged to stay financially viable. However, because these strategies were not enough to counteract the consistent erosion of profit margins after BBA, Congress eventually responded by increasing reimbursement to hospitals. The impact of these policies on quality of care was mixed, although overall studies found increased readmission rates, increased mortality, and reduced nursing staff. It is important to note, however, that the goal of these cost containment policies was not to increase quality of care. More recent policies have attempted to encourage hospitals to improve by requiring public reporting of both process and outcome measures and the use o f pay for performance. The impact of these new policies is mixed. The general consensus is that the use of public reporting and pay for performance to influence adherence to process measures are effective as long as the financial incentive is substantial, but the impact on patient outcomes is inconclusive. payment policy, which has both cost containment and financial incentive components, on patient outcomes and hospital financial performance? Would it have the int ended effect of improving patient outcomes? And how would it affect the financial performance of general short term acute hospitals? The next chapter will develop a conceptual framework for answering these questions.

PAGE 37

37 CHAPTER 3 CONCEPTUAL FRAMEWOR K payment policy affects patient outcomes and hospital financial performance. This chapter first reviews theories linking financial incentives to patient outcomes and financial performance. Then b ased on the theories, frameworks will be developed to predict the impacts of the non payment policy on patient outcomes (quality) and on financial performance. Economic theory suggests that, under certain conditions, the competitive model of firm behavior optimizes social welfare. However, Arrow (1963) pointed out that the medical care industry does not meet the conditions for a competitive market. In particular, five characteristics, which he summarized jointly as uncertainty, result in a non competitive industry. These characteristics are the irregularity and unpredictability of the demand for medical care, the expected behavior of physicians (the ethical restrictions on the activities of the physician as a seller), the uncertainty of the product (includi ng informational asymmetry), supply conditions (restrictions to entry in the medical profession due to licensing and the high cost of medical education), and pricing practices (including extensive price discrimination by income and contractual arrangement s that bind patients to particular groups of physicians). According to Arrow, attempts to minimize these uncertainties have led to the special structural characteristics of the industry. For example, the uncertainty regarding medical needs and cost and th e product itself might be responsible for the direct involvement of third parties like the insurance companies and regulators, especially the government, in the medical care industry. Since the passage of Medicare and Medicaid, the government has become a significant provider, regulator, and financier in the industry. Several Medicare policies have been implemented over the years with the aim of improving social welfare. Medicare non payment

PAGE 38

38 for hospital acquired infections is one such policy aimed at in fluencing hospital behavior with the aim of improving patient safety. Special characteristics of the medical care industry and the plurality of stakeholders suggest that any single theory will not be sufficient to predict the impact of a policy on hospita l behavior. Several different theories have been proposed to understand the impact of policies on hospital behavior. These theories span many disciplines including organizational theory, economics, sociology, anthropology, and psychology, with the objectiv e of expanding our understanding of the behavior of hospitals. In order to assess the impact of the Medicare non payment policy on patient outcomes and hospital financial performance, this chapter will develop two conceptual frameworks drawing on work rela ted to the quality of medical care, financial performance, financial incentives, and program evaluation. Framework Linking Financial Incentives to Patient Outcomes payment policy will affect pat ient outcomes. In parti cular, the Financial Incentive Structure Process Outcome Model is derived from theories of organizational behavior, economics, and program evaluation, as described below. Donabedian Structure Process Outcome Model Arguably the most cited model for evaluating quality of medical care is the Donabedian (1966) model, which has three dimensions to evaluating quality of care. The first is structure: you evaluate the quality of medical care by assessing the structures in place. Structure h ere includes the facilities, the competence of the medical staff in terms of their qualification, as well as the organizational and administrative structures. The second dimension is process: the process by which medical technology is used to achieve resul ts. Process defines what is actually done to the patient; for example, medications

PAGE 39

39 given, surgical procedures performed, recommendations given, and catheterization after surgery. Medicine is evolving and better ways of practicing medicine are constantly un der research. Process therefore evaluates the extent to which current knowledge is implemented to achieve quality. The third dimension is outcome: outcome is what happens to the patient. Although factors other than medical care influence the outcome of int erest, health outcome is acknowledged to have the best validation of the effectiveness and qual ity of medical care (Donabedian 2005). Empirically, researchers have studied the relationships among structure, process, and outcome. For example, Kunkel et al (2007) assessed how structure, process, and outcome can be used to analyze quality systems in hospitals. The study used survey responses of 386 hospital department heads or their quality control specialists in Sweden in 2004, defining quality systems as q uality assurance or continuous quality improvement programs. Using the SPO model shown in F igure 2 1, the study used both exploratory and confirmatory factor analysis and found that structure was strongly related to process with a factor of 0.72. Similarl y, structure was strongly related to outcome with a factor of 0.60, while process was positively related to outcome, given structure, with a factor of 0.2. A review article by Zinn and Mor (1998) reported on the structural attributes that affect outcome, w hich include volume of operation, hospital size, organizational factors (both internal and external), and managerial factors. The study reviewed articles that measured outcomes as death (mortality), disease (morbidity), disability (activities of daily livi ng scores), discomfort (pain), and dissatisfaction. Interestingly, ownership, system membership, and teaching status were not significantly related to outcomes

PAGE 40

40 The following structural attributes were significantly related to outcome: registered nurse to patient staff ratios (lower mortality, and better physical functioning in nursing homes), public hospitals (have higher mortality but mortality in private hospitals were mixed), higher volume facilities (had lower mortality rates), and payer mix (faciliti es with higher proportion of Medicaid recipients have higher pressure ulcer rates in nursing homes). The authors argued that looking at structure in isolation is insufficient to explain outcome differences between hospitals and that contingency theory is a useful addition in that strategies adopted by hospitals are contingent on external pressure, corporate and organizational culture, and therefore no one strategy could be said to be the standard way to achieve a desired outcome. Other studies focused on th e effect of volum e on hospital outcomes. Luft, Bunker, and Enthoven (1979) used data for short term general hospitals in the United States during 1974 and 1975 to examine mortality rates for open heart surgery, vascular surgery, total hip replacement, and cholecystectomy. The study found that hospitals that performed more than 200 surgical operations a year had 25 to 41% lower mortality rates than those performing less than 200 per year. Although not all results were statistically significant, there was a g eneral negative correlation between the volume of procedures done and the mortality rate. Taking a somewhat different approach but u sing the same data as Luft, Bunker, and Enthoven (1979). Shortell and LoGerfo (1981) found that hospitals with a higher vol ume of physicians under contract had lower mortality rates. The study also found that hospitals with lower mortality rates had high levels of involvement of medical staff in the hospital governing board, frequent meetings of medical staff committees, and a higher percentage of active staff physicians on contract.

PAGE 41

41 Nurses play an important role in patient care and several studies have looked at the impact of various nursing measures on quality outcomes. Kane et al (2007) in reviewing observational studies i n the literature, found that higher registered nurse staffing was associated with less hospital related mortality, hospital acquired pneumonia, and other adverse patient outcomes like cardiac arrest and failure to rescue. Aiken et al (2003), using dischar ge data from 168 non federal general hospitals in Pennsylvania, studied the association between the educational level of nurses and patient outcomes and found that, holding other characteristics constant, a 10% increase in the number of nurses with a bache decrease in mortality and failure to rescue rates of surgical patients. or reducing hospital acquired infections) will be influenced by hospital structure and processes. Key structural characteristics to be considered are volume of operation, physicians under contract, nurse staffing, and nurse educational background. Important process measures, discussed in the introduct ion, include having an infection control practitioner on staff, using surveillance with feedback, using coated catheters and giving antibiotics prophylactically. However, the SPO model does not address the impact of financial incentives on the quality of m edical care. Other theories as described below will be used to modify this model to include financial incentives. Principal Agent Theory Economic theories have played a leading role in explaining the impact of financial incentives on behavior. One of the t heories often used is the principal agent theory. In this theory the principal is the one who employs or contracts with an agent to perform some form of work on

PAGE 42

42 pri ncipal (for example, effort), and therefore financial incentives are used by principals to influence perfor mance of the agent (Prendergast 1999). The use of financial incentives is common in the business world. For example, executive compensation schemes are justified by boards of directors as a means of getting chief executives and managers to give of their best effort. Such incentives have come in various forms, especially as bonuses or stock options. Numerous studies suggest that such arrangements have influenced the performance of agents. In one such study by Prendergast (1999), performances linked to piece rates were higher than those linked to fixed salaries, with productivity increasing by 35% for workers fixing windshields when the payment structur e changed from fixed salary to piece rate. In the health care sector, DeBrock and Arnould (1992) used this theory to explain the contractual arrangements between HMOs and physicians. The principal in this case is the HMO which contracts with physicians (ag ents) to take care of the health needs of their members. Because of information asymmetry, HMOs have used several payment arrangements to influence the behavior of physicians. The principal (HMO) devises a payment scheme to maximize profit predicated on be tter treatment of its members by physicians including efforts to control cost. The agents (physicians), on the other hand, try to maximize their expected utility, which depends on the payments received and the effort they devote to the task. The authors u sed financial, utilization, and contractual data for all HMOs in Illinois from 1985 to 1987. Using two utilization measures (admissions and number of visits) as dependent variables and the various payment arrangements as the primary independent variables, the study found that utilization was higher for physicians on fee for service arrangements than for those on capitation and that utilization was 16.4% less for arrangements that targeted individual rather than group behaviors.

PAGE 43

43 Schneider and Mathois (2006) also used the principal agent theory to predict that insurers actively engaged in monitoring costs would have a lower health care utilization by its members compared to those who are passive about it. The study, using the Medical Expenditure Panel Survey d ata for 1996 and 1998, found that length of stay was lower at practices where HMOs actively monitored costs. payment policy is intended to influence the efforts hospitals make to prevent hospital acquired infections. In this case Medicare i s the principal and can influence its inpatient reimbursements. That is, hospitals are more likely to invest time, effort, and money into their structures (e.g. h iring qualified infection control professionals) and processes (e.g. improved surveillance systems) in order to improve outcomes (reduce hospital acquired infections) if Medicare reimbursement is a significant portion of their inpatient revenue. Financial incentives will therefore influence outcome either through structure, process, or both. Production of Quality Another model useful in developing the conceptual framework is a microeconomic model of the production of quality by a health care organization (A very and Schultz, 2001). From a under a preference structure shaped by professio 266). That is, any effort to increase the pr oduction of quality beyond professional obligations and altruism will depend on the financial returns the provider anticipates. The study then looked at how policies using pay for performance to institute penalties to meet standards would influence the mod el. the marginal return from quality equals the margina 269). Regulatory demands increase the cost of production and therefore providers would respond to

PAGE 44

44 financial incentives (including penalties) if the marginal return is at least equal to the marginal cost of producing the desired level of quality. Thus financial incentives due to regulation or policy will again influence outcom e through either structure or process adjustments. Effectiveness of Financial Incentives A final contribution to the conceptual framework comes from an article by Young et al (2005) intended to aid the design and implementation of pay for quality progra ms. The article lists the following five key issues that are essential for maximum impact of financial incentives: Awareness of quality incentives Knowing about the incentive and what the target and expectations are influences the outcome. For example, a study of Medicaid HMO enrollees by Hillman et al (1998) randomized physicians into intervention and control groups and offered 10 to 20% of capitation payments to the top 6 practices with the highest screening rates for eligible members for breast, cervi cal, and colorectal cancer. The study, conducted from 1993 to 1995 with a Medicaid managed care organization in Philadelphia, found that, although screening rates doubled overall, there was no significant difference in the screening rates between the inter vention and control groups, suggesting that the financial incentive did not influence the some physicians in the practices about the existence of incentives. In the case of Medicare non payment for hospital acquired infections however, this factor is unlikely to play a role given that the policy change has been widely disseminated. Size and Structure of the incentive The size and the form of a financial incentive, whether it is a bonus or a holdback, will influence the outcomes. For instance, in the study by Hillman et al (1998), the authors also suggested that the insignificant difference between the intervention and control groups was due to the smallness of the incentive. On the average, practices received $775.00 per site which the authors suggest was not enough to incentivize the

PAGE 45

45 physicians. As a p oint of comparison, the pay for reporting program that CMS implemented through the Premier Hospital Quality Incent ive Demonstration (between the fourth quarter of 2003 to the third quarter of 2006) paid 2% to the top 10% of achievers and 1% to the next 10%. CMS paid $8.85 Million to 123 hospitals in the first year of the demonstration ($680,769.00 per hospital on aver age), $8.7 million in the second year to 115 hospitals ($695,652 per hospital on average), and $7 million in the third year to 112 hospitals ($625,000 per hospital on average) (Rand 2010). These amounts were substantial enough to elicit responses from hosp itals, given the significant increase in adherence to process measures as reported by the project. The payment policy will be influenced by the size of the revenues they may lose from this p olicy. Accountability for quality targets The extent to which individuals/ teams are held responsible for achieving targets will influence outcomes. For example, in the study by Hillman et al (1998) single practice physicians outperformed group practices. In the case of this study, CMS is not targeting individual physicians but the hospital as a unit, therefore the impact will depend on the extent to which hospital administrators can rally clinicians to buy into whatever program they intend to put in place to mitigate the impact of the policy. Hospitals in systems with central control and strong clinician influence in the administration of the facility might outperform their counterparts. The clinical relevance of the quality targets Clinicians are more lik ely to respond to financial incentives if they believe that the quality targets have clinical significance. The IOM and others have made the case that patient safety can be improved with a systems approach. It is unlikely that a clinician will dispute the importance of preventing nosocomial infections. The

PAGE 46

46 onus rather lies more with hospital administrators the extent to which they are willing to invest in order to improve quality indicators. Fairness of the pay for quality programs The quality targets set in pay for quality programs are very important to providers. Targets include thresholds, percentage improvements, and ranking of providers. Ranking is often criticized because it does not motivate providers who do not believe they can make it to the top to even try. Medicare non payment policy is not based on any of these targets but aims at zero tolerance and therefore hospitals that are already actively and successively controlling hospital acquired infections are less likely to suffer reimbursement losse s. Adherence to infection control guidelines is costly, and hospitals do not have much control over which patients get admitted in the communities they serve. Certain kinds of patients are more prone to infections and therefore different hospitals are like ly to respond differently based on their levels of capital and expertise. Larger hospitals, specialty hospitals, and academic medical centers are more likely to outperform smaller and rural hospitals due to size and expertise. Additionally, certain patien t characteristics have been shown to empirically influence outcome and have been used to risk adjustment outcome measures. These characteristics include age, gender, race, severity of condition, and type of insura nce a patient carries (Hiestand et al. 2004 ) Combining the theories described above yields a model incorporating financial incentives in the standard SPO model As shown in Figure 2 2 environmental factors like regulations or policies can influence the patient outcomes of hospitals either through the structure of the hospital, the processes in place, or both. For this study the policy of interest is Medicare non payment of hospital acquired conditions. The non payment policy will result in financial pressure on hospitals, the size of

PAGE 47

47 which could in fluence the structural characteristics of hospitals and could influence patient outcomes either directly or through the improved processes implemented by hospitals. Some of the structural changes that could result from this policy include change of nursing mix or number of nurses per patient days, or greater role of physicians in the management of the hospital. Process changes could be improved infection control programs, hiring of an infection control specialist, adoption of surveillance systems from other successful hospitals, or increased adherence to proven infection control processes. The patient outcomes of interest are primarily the reduction or elimination of catheter associated urinary tract infection, vascular catheter associated infection, mediast initis, and secondarily the reduction or elimination of the other five hospital acquired conditions targeted by Medicare (decubitus pressure ulcers, foreign body left in the body, air embolism, blood incompatibility, and falls & fractures, dislocations, in tracranial injury, & burns) In summary, the probability of a decline in hospital acquired infections is a function of the size of the financial incentive, patient characteristics, and the structure and process characteristics of the hospital Framewor k Linking Financial Incentives to Financial Performance the financial performance of hospitals. Why do we need to know the impact of the policy on the financial performance of hospitals? Because past financial performance is a good indicator of an depends on its financial performance. A study by Hernandez and Kaluzny (1983) included a s urvey of 231 hospitals that closed or relocated from 1975 to 1977. Of the problems cited for the closures, 26.8% were due to financial problems, 14.3% were due to low occupancy, 23.4% for

PAGE 48

48 replacement with a new facility, and 10% due to lack of medical staf f. Financial health thus played a significant role in most closures. Neoclassical Theory of the Firm maximize profit. Profit maximization occurs when the marginal cost is equal to the marginal revenue. Although the neoclassical theory of the firm has been criticized on the grounds that some firms have different objectives such as increasing in size or market share, environmental and social objectives, profit (the difference between total revenue and total cost) is essential to a challenged as a narrow view because hospital executives are influenced by several stakeholders. Public regula tions necessitate admission of the uninsured under certain conditions, some hospitals are sponsored by religious associations with a defined mission, and some hospitals depend on philanthropy so they serve their communities in ways that are not consistent with economic principles of profit maximization. Because the hospital sector is dominated by non profit organizations, alternative economic theories of nonprofit hospital behavior such as those proposed by Newhouse (1973) and Hoerger (1991) would seem more appropriate. Bazzoli et al (2004 quantity and/or quality of services they produce subject to a break even constraint (i.e. revenues equal costs) or, perhaps more r ealistically, a target profit level needed t o maintain future 402). Thus when analyzing hospital behavior, particularly that of nonprofit hospitals, the assumption is that hospitals will strive to maximize utility, which may depend on missi on and other non hospital behavior, particularly for hospitals having a substantial nu mber of Medicare patients.

PAGE 49

49 Factors Affecting Hospital Financial Performance Gapenski, Bruce, and Langled Orban (1993) studied the determinants of hospital profitability, grouping them into four categories: organizational variables, managerial variables, pa tient mix variables, and market variables. The study used data for 169 general acute care hospitals in the state of Florida in 1989. Of the organizational variables, teaching status (negatively), hospital size (negatively) and ownership (for profit more pr ofitable) were found to be associated with profitability. Ownership has received a lot of attention in the literature with mixed results regarding the effect on profitability (Tennyson 2000 ; Younis 2006 ). Teaching hospitals normally have higher costs becau se of the types of cases they end up with, but they also normally get higher reimbursement due to case mix adjustments (Clement et al. 1997). System membership also has been found to have mixed effects on profitability. Bazzolli et al (2000), using 1995 M edicare cost report and American Hospital Association (AHA) data, found that the impact of systems on financial performance depended on the level of centralization of control of performed better than those in contractu (p 234). Clement et al (1997), using Medicare cost report for 1994 and 1995, also found that hospitals in systems had higher net revenues but not better cost efficiency. Of the manageri al variables in the study by Gapenski Bruce, and Langled Orban (1993), service index and labor intensity are the two relevant to this study. Service index was measured by the number of services offered in the hospital, while labor intensity was the number of full time employees (FTE) per patient day. Service index was positively associated with profitability, while as expected labor intensity was negatively correlated with profitability. Another study used Medicare cost reports from 1997 to 2004 and found that higher RN wages were associated with lower revenue, while the effect of RN wages on profit margin was not statistically significant

PAGE 50

50 (Schneider et al. 2007), and Tennyson (2000), using data for Florida hospitals in 1986 and 1992 found that higher staf f ratios were associated with lower financial performance. Of the patient mix variables, Gapenski Bruce, and Langled Orban (1993) found only Medicaid and Medicare mix to be significantly negatively associated with profitability, with the results not being statistically significant for case mix, average length of stay, outpatient mix, or managed care mix. Sear (2004) analyzed financial data for 25 hospitals in Tampa, Florida from 1990 to 2001 and found that commercially discounted payer mix was positively a ssociated with financial performance. Finally, of the market variables, Gapenski Bruce, and Langled Orban (1993) found area wage rate (negatively), physician density (positively), and patient income (negatively) measured as per capita income to be signifi cant predicators of profitability. Contrary to predictions, however, the study did not find that competition, measured by hospital concentration, had a significant effect on profitability. The Profitability Model Financial performance is a broad term an d can be measured in a variety of ways; some measures assess long term viability of a firm while others focus on the short term. This study will examine short term effects of the non payment policy and therefore focuses only on one aspect of financial perf ormance profitability, defined as the difference between total revenues and total costs. As discussed in the literature review, the most significant reaction of hospitals to financial reform changes like prospective payment and the Balanced Budget Act of 1997 was cost payment policy, as discussed earlier, the impact will depend on the extent to which the policy will affect the revenue of hospitals. Hospitals might either further reduce cost of production to offset t he reduction in revenue or invest in more

PAGE 51

51 sophisticated systems to further reduce the likelihood o f nosocomial infections (Feder et al. 1987). As pointed out earlier, hospitals that have been successful in reducing infection rates used strategies that woul d tend to increase cost, such as increased surveillance, other preventive measures like the use of coated catheters, or i ncreased labor (Apisarnthanarak et al. 2007). Hospitals spend a lot of resources to treat nosocomial infections, and it is estimated that they spend more tha n they get reimbursed (Anderson et al. 2007). Therefore, the higher the nosocomial infection rate the higher should be the inpatient care cost per patient discharged. The impact of the CMS policy on hospital cost, however, will de pend on the strategy adopted by the hospital. Hospitals that do not invest substantially to augment their infection control programs will have no statistical difference in their patient care cost, while hospitals that do invest in reducing infection rates due to this policy are likely to have higher cost in the interim but lower infection post policy. Revenue is the other key determinant of profitability. Schneider et al (2007) describe a d other explanatory variables that affect revenue. The Medicare non payment policy is designed to reduce output price for hospitals with nosocomial infections. Reimbursement is based on the DRG payment system and before the policy Medicare had paid for hig her coded DRGs which had included nosocomial infections. Under the new policy hospitals will not be able to code such procedures at a higher conditions post policy. So me evidence in the literature suggests that hospitals respond to such reimbursement losses by either charging more to other payers through contract negotiations or increased outlier claims (Sear 2004). Therefore hospitals with previous higher infection rat es and a higher proportion of Medicare patients are likely to have reduced revenues.

PAGE 52

52 The financial impact of the Medicare non payment will depend on behavioral changes adopted by hospitals. Behavioral changes will depend on the financial pressure hospitals perceive will result from the implemented policy. That perception would be guided by the number of Medicare patients they have, their current financial status, and their historic infection rates. Hospitals with higher financial pressure are more likely to incur higher patient care cost (because of initial infection control costs), lower revenues (because of the number of Medicare patients they have) and lower margin (because margin is the difference between revenue and cost). In summary, profitability w ill be a function of financial incentive, organizational variables, managerial variables, patient mix variables, and market variables. Hypothesis Based on the conceptual framework and research questions described above I propose the following hypothese s g rouped according to the research question; A. Patient outcome question 1. Medicare non payment policy will reduce the probability that a n admitted patient acquires any of the hospital acquired conditions. 2. The reduction in probability will be higher for hospit als with higher previous year proportion of Medicare admissions. 3. The reduction in probability will be higher for hospitals with lower previous year operating margins per admission. B. Financial performance question Medicare used to pay for the hospital acq uired conditions but the literature suggests that hospitals incur additional cost (AHRQ, 2010) in treating patients with these conditions therefore; 1. The higher the rate of hospital acquired conditions the higher the operating cost of hospitals.

PAGE 53

53 2. Since Medic are used to pay higher DRGs for these conditions and other third party payers have not announced cessation of these payments, the higher the rate of these hospital acquired conditions, the higher the operating revenues of hospitals. 3. There will be a reducti on in inpatient care revenue post policy the higher the number of hospital acquired conditions not paid for by Medicare. 4. The reduction in inpatient care revenue will be higher for hospitals with higher proportion of Medicare admissions 5. There will be an i n crease in inpatient care expenses post policy the higher the number of hospital acquired conditions not paid for by Medicare. 6. The incre ase in inpatient care expenses will be higher for hospitals with higher proportion of Medicare admissions. 7. Profit margin s will be lower for hospitals with higher proportion of Medicare admissions post policy and lower for hospitals with higher rate of hospital acquired conditions not paid for by Medicare post policy Figure 3 1 Kunkel Structure Process Outcome Models

PAGE 54

54 Figure 3 2. The Financial Incentive Structure Process Outcome model

PAGE 55

55 CHAPTER 4 METHODOLOGY This chapter describes the study design, data, and variables used to answer the two research questions addressed in this dissertation. The first question is to assess the impact of to assess the impact of the policy on hospital financial performance. Study Design The study is a po licy evaluation that uses a before after design to assess the impact of the policy on patient outcomes and hospital financial performance. The federal fiscal year starts October 1 st of the previous year to September 30 of the year in question. This policy affects Fiscal Year 2009, going into effect on October 1, 2008. Any claims submitted by a hospital dated October 1 st 2008 and later are subject to the non payment of hospital acquired condition clause of the policy. Timeline for the study : Pre policy Jan 1, 2007 to Sept. 30, 2008 Post policy Oct. 1, 2008 to Dec. 31, 2009 The dissertation compares the pre policy period (January 1, 2007 to Septem ber 30, 2008) to the post policy period (October 1, 2008 to December 31, 2009). Because the policy was impleme nted so recently, only a year of data are available for the post policy period. The dataset used have all variables needed for the study only from January, 1 2007. Consequently, the results of this study must be considered preliminary, given that they refl ect the impact of the policy only one year after implementation. Overview of All Data Sources There are no publicly available datasets on patient outcomes at the national level to conduct a post policy analysis based on the specifications of this study. Pr evious studies have

PAGE 56

56 used simulations or projections both at national (McNutt et al. 2010) and state (McNair Luft, and Bindman 2009) levels using data from the pre policy years. However, the state of Florida makes available hospital discharge data that in clude information on patient outcomes and whether the condition was present on admission or not this dataset was used to answer the question on patient outcomes. The state of Florida also collects utilization and financial data for all hospitals and this d ataset was used to answer the financial performance question. The Area Resource File, process measures reported on the Hospital Compare website, data from American Hospital Association, were used to augment the main data sources. Florida Hospital Uniform Reporting System Inpatient data Since 1988 the Florida Agency for Health Care Administration (AHCA) has maintained an inpatient discharge dataset with data on approximately 2 million patients per year. More recently, the dataset includes 2,563,643 discha rges in 2007, 2,571,688 in 20 08, and 2,491,033 in 2009 (AHCA 2010). The data set includes all hospitals in Florida with the exception of VA hospitals. The dataset also includes the following patient level characteristics: demographic characteristics (e.g. age, gender, and race), admitting diagnosis, principal diagnosis present on admission indicator, principal procedure code, length of stay, a 3 digit number representing the assigned Medicare severity diagnosis related group (MS DRG), charges, and payer ty pe. This dataset provides the patient demographics, diagnosis, and was used to identify patients that had nosocomial infections present or absent on admission. Financial data AHCA is also the source of the financial data for the state level analysis. The Florida Hospital Uniform Reporting System collects data yearly for all hospitals in the state. The dataset contains hospital financial data (e.g. income and balanced sheet statements) audited

PAGE 57

57 by a Florida licensed certified public accountant as well as inf ormation on the type of hospital (long term, short term), type of control (investor owned, government or non profit), whether it is a major organ transplantation hospital, and other hospital characteristics. This dataset was used to obtain variables such a s inpatient care revenues, inpatient care expenses, operating margin, and hospital characteristics like the hospital ownership and size. Area Resource File The Area Resource File (ARF) from the Health Resources and Administration Services (HRSA) was used t o provide county level information (e.g. the number of physicians and population density in a given metropolitan statistical area ). The dataset includes information about utilization and other characteristics of health care facilities. Hospital Compare Thi s is a website maintained by CMS, where process measures reported by hospitals to Medicare are made public so that patients can compare the performances of hospitals. This dataset was used to obtain process scores of Florida hospitals America Hospital Ass ociation D ataset This dataset was used to obtain variables on nurse full time equivalents for both registered nurses (RNs) and licensed practical nurses (LPNs). Hospital Blue Book This reference book contains information on hospital characteristics includi ng names of hospital executives and whether they have an infection control specialist on staff Empirical Specification Research Question #1 Based on the con ceptual framework described in C hapter 3, the empirical specification for the first question is:

PAGE 58

58 Probability of hospital acquired infection/ condition process, and patient characteristics) The following section d escribes the variables that were used to measure the various factors in the empirical specification ( details on variable measurement and data source will be provided in a subsequent section). Hospital Acquired Infections The non payment policy targets three hospital acquired infections for 2009: urinary tract infections due to catheterization (CAUTI), vas cular catheter associated infection (VCAI) and mediastinitis (SSI) a surgical site infection after coronary artery bypass graft. The probability that a patient with any of the conditions listed above acquired it while in the hospital is the patient outcom e of interest. Financial Incentive The non payment policy is designed to act as a negative financial incentive for hospitals. The incentive here is the loss of revenue due to a Medicare patient acquiring any of the hospital acquired conditions which befor e implementation of the policy were reimbursable. The financial incentive was captured by a time variant distinguishing between time before policy and time after policy, pr a variable capturing the financial state of a hospita l before the commencement of a financial year, and a continuous variable capturing the extent to which a hospital depends on Medicare Structure Structural characteristics were captured by hospital characteristics, including the size whether it is a teac hing hospital the number/mix of nurses and the type of ownership Process Measures used to capture process include the number of surgeries and catheterizations preformed by a hospital and whether they have an infection control specialist on staff.

PAGE 59

59 Other covariates Other c ovariates include patient characteristics such as age, gender, race, the number of co morbidities and whether they used the ICU during their stay. For Research Question #1, the final version of the empirical specification thus is; Proba bility of hospital acquired infection (CAUTI, VCATI, SSI) structure process and other covariates ] Research Question #2 For the second research question the empirical specification derived from the conceptual framework is: Financi variables, patient mix variables, market variables, and other covariates). Financial Performance Financial performance was measured by three financial indicators; operating reven ue operating cost and operating margin as described in detail below. F INANCIAL INCENTIVE Financial incentive, the main independent variable, was be measured by the proportion of admissions t hat are Medicare patients, a time variant to distinguish betwee n time before policy and af ter policy, and the proportion of hospital acquired conditions not present on admission. O RGANIZATIONAL VARIAB LES Variables to measure hospital organizational characteristics include: OWNERSHIP (for profit (FP) and not for profi t (NFP) status), TEACH, and BEDSIZE. M ANAGERIAL VARIABLES Hospital managerial variables include PPERS, PSUPP, OCCRATE, PRICE, and TRANS. P ATIENT MIX VARIABLES H ospital level patient mix variables include CASEMIX, ALOS, and MCAID

PAGE 60

60 M ARKET VARIABLES Market v ariables, measured at the MSA level, include INCOME, UNEMP, POPDEN, MDPOP and HHI. For Research Question #2, the final version of the empirical specification thus is: Financial performance (REVENUE, COST, and MARGIN) /RHAC MC ARE YEAR ), organizational variables (OWNERSHIP, TEACH, BEDSIZE, R URA L ), managerial variable s (PPERS, PSUPP, OCCRATE, PRICE, TRANS), patient mix variables (CASEMIX, ALOS, MCAID), and market variables (INCOME, UNEMP, POPDEN, MDPOP, HHI)]. Variable Descript ions Dependent Variables Dependent variables for the patient outcomes question Here the patient outcome of interest is the probability of a patient with any of these conditions acquiring them w hile o n admission at the hospital. The three infections identif ied by Medicare for non payment ; catheter associated urinary tract infections (CAUTI), vascular catheter associated infection (VCAI), and mediastinitis (SSI) after a coronary bypass graft, were identified in the inpatient dataset using ICD 9 codes. The probability of infection is measured by the odds of a patient with any of these conditions but not present on admission. Cases will be the number of patients who were admitted without the infection but developed the infection after 48 hours in the hospita l (CAUTI/VCAI/SSI). Medicare requires hospitals to indicate whether the infection was present on admission (POA). For example a case is a patient without a POA indicator but with a diagnosis of either CAUTI /VCAI or SSI Controls are patients with any of t he conditions listed above but whose conditions were present on admission (POA). CAUTI s are the most voluminous of the three hospital acquired conditions so model 1 will look at the impact of the policy on

PAGE 61

61 CAUTI alone, model 2 will consider all three condi tions together as hospital acqui red infections (HAI) and a model 3 will look at all 8 conditions Medicare is not paying for as hospital acquired conditions (HAC) all 8 conditions will be defined in the next section. 1. Model 1 C AUTI is a dummy variable, =1 if a patient with CAUTI has a POA indicator marked N (not present on admission), else =0 (a patient with CAUTI with a POA indicator marked Y CAUTI was identified in the datas et with ICD 9 CM code of 996.64 (McNutt, 2010) 2. Model 2 HAI is a dummy variable, = 1 if a patient had either CAUTI, mediastinitis (SSI), or vascular catheter associated infection (VCAI) that was not present on admission, else =0 if those conditions were present on admission. Medicare will not pay for only SSI due to mediastinitis after c oronary artery bypass graft (ICD 9 CM 519.2 and procedure codes 36.10 36.19).(McNutt, 2010). VCAI is identified by ICD 9 CM 999.31 3. Model 3 HAC. The primary aim of this study is to examine the impact of the policy on nosocomial infections targeted by Medica re. However another important question is the impact of the policy on all hospital acquired conditions (HAC) targeted by Medicare in 2009. Patients who had any of the conditions listed below that was not present on admission will have HAC=1 otherwise HAC=0 a) foreign object retained after surgery (ICD 9 CM code 998.4/998.7), b) air embolism (ICD 9 CM code 999.1), c) blood incompatibility (ICD 9 CM code 999.6) d) stages 3 and 4 pressure ulcers (ICD 9 CM codes 999.6), e) hospital related falls/trauma (ICD 9 CM codes 8 00 829, 830 839, 850 854,925 929, 940 949, 991 994. f) UTI (ICD 9 CM code 996.64) or g) SSI (ICD 9 CM code 519.2 and procedure code 36.10 36.19) (McNutt,2010) Dependent variables for the financial performance question 1. DIREV Total inpatient care revenue per hosp ital per year. 2. REV Net operating revenue per hospital per year 3. DICOST Direct inpatient care expenses per hospital per year. 4. COST Net operating expenses per hospital per year 5. MARGI N O perating margin rate, which is (REV COST)/ REV per hospital per year. 6. OPMA RG Operating margin, which is (REV COST) per hospital per year

PAGE 62

62 Independent Variables For the patient outcome question: 1. POLICY as described earlier, POLICY is a dummy variable that differentiates between pre and post policy periods POLICY=1 for period af ter the policy (October 1, 2008 to December 3 1, 2009) and POLICY=0 for period before the policy (January 1, 2007 to September 3 0, 2008). 2. OPMPA this is a continuous variable of the admission per hospital. It is hypothe sized that hospitals with higher previous operating margins will be less likely to be under pressure to respond to this negative financial incentive from the policy than hospitals with higher operating margin It is assumed that hospitals strategize at the end of the financial year based on their income statements and balanced sheets and therefore hospitals with negative or close to zero operating margins will have a higher incentive to seek ways of improving their financial health. The policy which has the potential to reduce inpatient revenues will therefore create the financial incentive forcing hospitals to take steps to curtail the potential revenue loss 3. MCARE this is a continuous variable of the proportion of admissions that are from patients whose p rimary insurance is Medicare. The non payment policy targets only Medicare patients therefore the financial pressure will also be influenced by the proportion of Medicare inpatients in the patient payer mix. For the financial performance question 1. Model 1. HAC the proportion of admissions with any of the eight hospital conditions whether present on admission or not. 2. Model 2. RHAC the proportion of HAC not present on admission. 3. YEAR this is a vector of variables of the year of admission, 2009 is the omitted v ariable 4. MCARE the proportion of Medicare patient admissions. Covariates For patient outcomes question : Structure measures SIZE H ospital size is a categorical variable of the number of be ds per hospital (<100(small), 100 200 (moderate), >200 (big)). Smalle r hospitals are less likely to have sophisticated infection control programs. JAHCO requires hospitals with at least 250 beds to have an infection control professional on staff.

PAGE 63

63 TEACH A teaching hospital is defined as a member hospital of the Council of Teaching Hospitals. This variable will control for case mix differences and differing objectives between teaching and non teaching hos pitals (Frick, Martin, and Shwartz 1985) NURSEFTE This is the number of nurse hours FTE (RNs, and LPNs) per admission per hospital. RNFTE This is the number of registered nurse hours FTE per admission per hospital. MEDDIR This is a dummy variable that represents the presence of a medical director or medical chief of staff as part of the top hierarchy of a hospital. OWNERSH IP Hospital Ownership is a variable to distinguish between missions and management styles. This is a vector of binary variables for non profit, for profit, and government (where government is the omitted reference variable). Process measures PROCESS T h is is a continuous variab le that will capture the extent t o which hospitals adhere to recommended infection control practices. This informatio n come s from Hospital Compare a quality reporting tool made public by Medicare to aid comparison of hospitals by p atients, based on surveys of patients about their quality of care experience at a hospital. At the website, there are three measures that are applicable to this study. PROCESS is calculated as the average of the scores from three questions: a. percent of patients given antibiotics one hour before surgery to prevent infection, b. percent of surgery patients given the right kind of antibiotics, and c. percent of surgery patients whose preventive antibiotics were stopped at the right time ( within 24 hours after surgery). SURGERY This is a variable that represents the proportion of total admissions that are surgeries performed by a hospital per year. This variable is to capture differences bet ween hospitals that specialize in surgeries based on the proport ion they perform. TOTSUR T his is also a continuous variable that looks at the impact of volume of surgery performed by a hospital on patient outcomes. It is to control for the role of experience in controlling infection. PROFESSIONAL T his is also a proc ess measure which will be a dummy variable indicating whether or not the hospital has an infection control professional or department. CDC found that hospitals with an infection control professional had lower infection rates (Scott, 2009). Data for this va riable will come from websites of the hospitals or from calling the human resource department for the information. CATH T he proportion of admitted patients who were catheterized during their stay (procedure code ICD 9 CM 57.94, 57.95). Catheterization inc reases the probability of infection and hospitals with higher number of catheterization are prone to higher infection rates.

PAGE 64

64 Other Covariate Measures AGE This is a continuous variable of the age of a patient at admission. GENDER This is a dummy variable, =0 if female and =1 if male. RACE This is a vector of binary variables for Blacks, Hispanics, and others (one variable for all other races) with Whites as the omitted reference group. LOSDAYS T his is a continuous variable of the length of stay of a give n patient. Patients are prone to acquire infections the longer they stay in the hospital. COMORB This is the number of co morbidities recorded for a particular patient. Co morbidities increases the chances for pat ients to get infected (McGiffin et al. 200 2). ICU This is dummy variable indicating if a patient was admitted to the ICU or not. ICU patients are more prone to getting hospital acquired infections. RURAL This is a dummy variable=1 for rural location and =0 for urban location. This will capture t he effect of the policy on rural hospitals. Because of the investments involved in infection control, rural hospitals might be slower to act as compared to urban hospitals. Covariates for the Financial Performance Question Organizational measures OWNERSHI P F or profit and not for profit status. For profit is the omitted variable. TEACH As defined earlier. TRANS This is a dummy variable (1, 0) that distinguishes hospitals with a major organ donation program in place from those without. BEDSIZE This is a discrete variable from 0 to 4. Hospitals with less than 100 beds=0, 100 to 199beds=1, 200 to 299 beds=2, 300 to 399 beds=3, and 400 and above=4. Managerial measure PRICE This is a continuous proxy variable representing the average input price prevailin g in a given MSA. This was obtained by dividing the sum of revenues of hospitals in an MSA with the sum of admissions of hospitals in the same MSA. OCCRATE T his is a continuous variable of the occupancy rate of a hospital. This was obtained by dividing to tal patient days (x 100) by the product of beds and number of days in operation (365). This is a volume measure and it is expected that the higher the occupancy rate the higher the profitability of a hospital.

PAGE 65

65 Patient mix measures CASEMIX Is the overall M edicare case mix per hospital. ALOS I s the average length of stay of a hospital calculated as the ratio of the total number of patient days divided by the total number of admissions. Hospitals with longer ALOS are more prone to worse patient outcomes (inf ections) (Dulworth and Pyenson 2004). MCAID T he proportion of Medicaid patient admissions. Previous studies show that it has negative impact on profitability. Market measures INCOME. I s a continuous variable of the per capita income of a given county in a year. MDPOP I s a continuous variable of the number of physicians per 1000 people in an MSA. The more physicians in a given county the higher the probability of physician office practices and hence higher hospital admissions. However that number could als o reduce outpatient admissions of hospitals since they will compete directly with free standing clinics of hospitals as well as reduce emergency room usage. POPDEN T his is a continuous variable of the population density of a given MSA. The denser an MSA, admissions in surrounding hospitals are more likely to have higher admissions. HHI I s a proxy for competition in a market calculated based on the market share of a given hospital in an MSA; the market share is the proportion of admissions of a hospital in a given MSA. HHI= sum of squared market shares of all hospitals in the county. The lower the value the more competitive the market. PPERS T his is the average salary or a wage per full time equivalent of a patient care worker PSUPP T his is the total sala ries of the support staff per bed Table 4 1 summarize s the variables described above per research question. Econometric Methods Clement et al (2007) examined the impact of the Balanced Budget Act of 1997 on hospital patient safety using a pre and post d esign. This study modifies this design to account for differences between hospitals in their ability and strategy to respond to policy changes. The primary goal is to look at the effect of the policy on patients, i.e., whether the policy has

PAGE 66

66 improved patie nt outcomes. The secondary goal is to assess how the changes, if any, vary among hospitals. Patient Outcome Model The analysis uses a Hierarchical Linear Model (HLM) to capture both levels of analysis. HLM is a statistical tool that accounts for the hiera rchical (multi level) nature of a given data. The patient outcome model has patient and hospital level data. HLM helps to analyze the data at these multiple levels to account for clustering due to hospitals. If aggregate values per hospital are used (e.g. infection rates per hospital for the dependent variable), patient level effects could be masked by these mean values. Using patient level data, we can find out the impact of the policy on the probability of a typical patient to acquire these target conditi ons in a typical hospital instead of the change in probability of a typical hospital. HLM will produce the extent of variability due to hospit al effects (Raudenbush and Bryk 2002). The model specification for the patient outcomes question is; Logit (INFECT ION it) 0 1 POLICY ht 2 IPRPA ht 3 MCARE ht 4 POLICY*IPRPA ht 5 POLICY*MCARE ht 6 SIZE ht 7 MEDDIR ht 8 RNFTE ht 9 NURSFTE ht + 10 PROFESSIONAL ht 11 COMORB iht 12 SURGERY ht 13 TEACH ht 14 CATH ht + 15 AGE iht 16 GENDER iht 17 RACE iht 18 RURAL ht 19 PROCESS ht 20 TOTSUR ht + 21 TOTCATH ht + 22 ICU iht 23 LOSDAYS iht + r it Where i is at the patient level, h is at the hospital level, and t represents year. The error term r it captures the hospital level characteristic s Financial Performance Model For the second question on financial performance, the focus is on hospital level data and OLS was a good for the data. Diagnostic tests including test for linearity, leverage, normality, multicollinearity were performed. A mix ed effect model using time as a repeated variable was

PAGE 67

67 performed using compound symmetry as well as auto regression covariance structure but the results were similar to OLS so the results reported are that of the OLS model. The model specification for the f inancial performance question is; (Y ht) 0 1 HAC ht 2 MCARE h t 3 TIME ht 4 HAC*TIME ht 5 MCARE*TIME ht + 6 MCARE*HAC*TIME ht 7 BEDSIZE ht 8 MDPOP m 9 INCOME m t 10 UNEMP m t + 11 CASEMIX ht 12 OWNERSHIP ht 13 POPDEN m 14 TEACH ht 15 ALOS h t 16 HHI m t + 17 PRICE mt 18 OCCRATE ht 19 TRANS ht 20 ADM ht + 21 PPERS ht + 22 SPSUPP ht ht Whe re h is at the hospital level, m at the MSA level and t is time. Where Y is the outcome of interest (IREV, REV COST, DICOST, OPMARG or MARGI N)

PAGE 68

68 Ta ble 4 1. Variables for the patient outcomes question Dependent variables Description Type Identification CAUTI Catheter associated urinary tract infections Dummy variable =1 if CAUTI not present at admission and =0 if present at admission Identifi ed by ICD code 996.64 HAI Hospital acquired infections CAUTI/vascular catheter associated infections(VCAI)/ surgical site infections(SSI mediastinitis only) Dummy variable=1 if HAI (any one of the 3) is not present at admission and =0 if present at admiss ion Identified by ICD codes CAUTI=996.64 VCAI=999.31 and SSI= 519.2 HAC Hospital acquired conditions HAI +foreign body/air embolism/blood incomp./pressure ulcers/falls & trauma D ummy variable=1 if HAC not present at admission and =0 if present at admissio n Identified by ICD codes. HAI=as above Foreign body=998.4 Blood incomp.=999.6 Pressure ulcers=707.00 707.09. Independent Variables Description Type Identification POLICY Time indicator D ummy variable=1 if post policy=0 pre policy Jan 2007 to S ept.30, 20 08=0 and Oct.1 2008 to Dec 2009=1 MCARE Proportion of Medicare inpatient admission C ontinuous variables Medicare admission/Total admission/hospital OPMPA per admission. C ontinuous FHURS. TREND Time indicator in quarters D iscrete 2009=12 Control variables Patient characteristics AGE D emographic C ontinuous Age at admission PAYER Source of payment by patient D ummy Medicare is the omitted variable COMORB # of other diagnosis recorded C ontinuous Sum of number of diagnosis recorded SEX D emographic D ummy Male=1 Female=0 RACE Demographic white, blacks, Hispanics, others D ummy Whites omitted ICU Intensive care usage D ummy =1 if used =0 if not used LOSDAYS Length of stay C ontinuous Admitted to discharge Pr ocess measures at hospital level PROCESS Proxy for adherence to guidelines C ontinuous From hospital compare SURGERY Number of surgeries / total admission /hospital/year C ontinuous Identified by operating room charges .Revenue codes 360 to 369 CATH # of catheterized patients per Total admission C ontinuous ICD codes 57.94/57.95

PAGE 69

69 Table 4 1. Continued Dependent variables Description Type Identification Structure measures SIZE Number of beds D ummy <100 (small),100 250 (moderate), >250 (big) abov e 250 omitted variable TEACH Teaching or non teaching hospital D ummy =1 for teaching hospitals =0 otherwise Independent Variables Description Type Identification MEDDIR Medical director/chief of staff D ummy Hospital blue book NURSEFTE Total FTEs of LP Ns and RNs/patient days C ontinuous From AHA data RNFTE Total RN FTEs/patient days C ontinuous From AHA data RURAL Location of hospital D ummy =1 for rural else 0 Table 4 2. Variables for the financial performance question Dependent variables Descriptio n Type Identification REV Total patient care revenues C ontinuous As per income statement DIREV Direct inpatient care revenues C ontinuous Inpatient revenue weighted by deductions COST Total patient care expenditures C ontinuous As reported in the income s tatement DICOST Direct inpatient care expenses C ontinuous FHURS MARG Operating margins per hospital per year C ontinuous As reported in the income statement of hospitals Independent variables POLICY Time indicator Dummy Variable=1 if post policy=0 pre p olicy Jan 2007 t0 sept.30, 2008=0 and Oct.1 2008 to Dec 2009=1 HAC Proportion of HAC/ total admission Continuous Variables Using ICD codes for HAC to identify cases/total admissions/hospital /year RHAC Proportion of HAC not present on admission C ontinu ous With the use of the POA indicator Control variables Hospital characteristics BEDSIZE Hospital bed size C ontinuous Fhurs MCARE Proportion of Medicare /Inpatients days C ontinuous Fhurs MCAID Proportion of Medicaid/inpatient days C ontinuous Fhurs T EACH Indicator of type of hospital D ummy =1 for teaching hosp and =0 otherwise OWNERSHIP Type of ownership FP/NFP D ummy For profit=1 not for profit=0 CASEMIX Overall case mix C ontinuous Fhurs

PAGE 70

70 Table 4 2. Continued Dependent variables Description Type I dentification ADM Number of inpatient admissions/year/hospital C ontinuous Fhurs OUTV Number of outpatient visits per hosp/year C ontinuous Fhurs OCCURATE Occupancy rate C ontinuous Fhurs TRANSPLANT Hospital provides transplant services? D ummy =1 if yes and =0 if not County characteristics POPDEN Population per sq mile C ontinuous Area resource file MDPOP Physicians per 10000 population C ontinuous Area resource file MSA characteristics PPERS Average MSA FTE of patient care personnel C ontinuous Fhurs PSUPP Average MSA FTE of support staff C ontinuous Fhurs PRICE Output price per county C ontinuous Using Total Revenue of MSA/admissions of MSA UNEMPLOY Unemployment rate C ontinuous Bureau of labor and statistics INCOME Per capita income C ontinuous Bur eau of economics HHI Indicator of hospital concentration C ontinuous Fhurs Sum of share^2 PPERS The average patient care salary per FTE C ontinuous FHURS PSUPP The average support staff salary per bed C ontinuous FHURS

PAGE 71

71 CHAPTER 5 RESULTS This chapter s ummarizes the results of the dissertation and it is divided into two parts. Part 1 discusses results of the patient outcome s analysis and P art 2 discusses the results of the financial performance analysis The discussions in each part include the des cripti ve statistics, ANOVA analysis and then the regression results Part 1: Patient Outcome Analysis Descriptive Statistics Over the period of analysis (2007 2009), t here were about 2.5 million discharges from acute care hospitals per year. A very small number of patients had conditions targeted by the policy ranging from 4,848 discharges with catheter associated urinary tract infections (CAUTI) before the policy to 72,998 discharges with at least one of the 8 targeted conditions before the policy. Of these, as shown in T able 5 1 the proportion of discharges with catheter associated urinary tract infections (CAUTI) not present on admission (NPOA) were 19% before the policy (January 1, 2007 to September 30, 2008) and 18% after the policy (October 1, 2008 to Dece mber, 31, 2009). Also, the proportion of discharges with at least one of the three hospital acquired infections (HAI) NPOA was 15% pre policy and 10% post policy Additionally, the proportion of discharges with at least one of the eight hospital acquired c onditions (HAC) in the Medicare policy NPOA was 16% pre policy and 15% post p olicy. Overall, about 1.8% of acute care patients discharged between 2007 and 2009 had at least one of the 8 hospital acquired conditions identified by Medicare. This is higher th an the 0.11% identified by McNair et al in the California data which might be because they used only Medicare discharge data The 1.8%, however, is lower than the 4% McNutt et al. (2010) found in his academic center sample. The

PAGE 72

72 variables below are divide d into 2 levels; level 1 is at the patient level and level 2 is at the hospital level From Table 5 1 of the about 5000 discharges with CAUTI, 19% were not present on admission before the policy decreasing to 18% after the policy. Similarly, 15% of hosp ital acquired i nfections were not present on admission before the policy falling t o 10% after the policy. Fin ally, 16% of all hospital acquired conditions were not present on admission before the policy decreasing to 15% after the policy. A typical patien t in the sample was between 62 and 72 years old and spent between 9 to 12 days in the hospital and had about 13 co morbidities. About 50 % to 60% discharges in the sample were male s with whites accounting for between 66 % and 77% of all discharges Between 3 8 % and 48% of discharged patients used intensive care Not surprisingly given the average age, about 80% of patients in the sample had either Medicare or Medicare HMO as the primary in surance, about 8% to 14% had private insurance and the rest had Medica id, self pay, or other government insurance s Table 5 2 describes the characteristics of hospitals from which these patients were discharged. For a typical hospital, about 50% of admissions were Medicare patients about 30% of admissions were surgery rel at ed, 41% of patients were catheterized during th eir admission stay, about 70% had an infection control specialist on staff and 77% had either a medical director or chief of staff as part of the executive staff. Additionally, 46% of hospitals in the sample have between 100 and 200 beds only about 15% have les s than 100 beds, a bout 3% are teaching hospitals and about 14% are located in a rural area. ANOVA A nalysis T able 5 3 summarizes the bivariate odds ratio of the dependent variables of interest comparin g them before and after the policy. The odds ratio compares the odds of a patient with

PAGE 73

73 CAUTI acquiring it on admission post policy with pre policy. For example, a patient with CAUTI post policy is 0.9185 times the odds pre policy meaning the patient is mor e like in the pre policy era to acquire CAUTI in the hospital than in the pre policy era. If the confidence interval includes 1 then the odds ratio is not considered statistically significant meaning there is more than 5% chance that the odds pre and post policy are the same. The odds ratios show a reduction in CAUTI, HAI, and HAC post policy but the reduction in CAUTI is not statistically significant because the confidence interval (0.8317 1.1015) includes 1. To further un derstand the differences among hos pitals with regards to these conditions, the analysis was extended to compare categories of hospitals based on their proportion of Medicare admissions per year. Based on the mean and standard deviations from a univariate analysis of the proportion of Medic are admissions per hospital per year, hospitals were classified into three groups; G roup 1 included hospitals with less th an 25% of Medicare admissions, G roup 2 included hospitals with Medicare admis sions between 25% and 50%, and G roup 3 comprised of hospi tals with more than 50% (the mean) of Medicare admissions. The graphs from Figure 5 1 onwards illustrate the differences between these categories of hospitals first across policy (pre (0) and post (1) ) and then across years (2007, 2008 & 2009). As shown in Figure 5 1 t he predicted mean (using proc mixed procedure) infection rate of CAUTI decreased significantly from about 0.1925 pre policy to about 0.1575 post policy for Groups 1 hospitals but infection rates in hospitals in Groups 2 and 3 remained about the same. C ompar ing the groups by year instead, as shown in Figure 5 2, the average infection rate of CAUTI decreased for all groups from 2007 to 2008 The decrease continued for Group 1, remained approximately the same for Group 2, and increased for Group 3 from 2007 to 2008

PAGE 74

74 These suggest that hospitals might have invested to reduce the infection rate prior to the implementation of the policy in October, 2008, but have not further reduced the rate since then. However hospitals with less than 25% Medicare admissions rather showed continuous decrease pre and post Policy while hospital with more than 50% Medicare admissions performed poorly post policy. The next two graphs illustrate the impact of the policy on HAIs among the three groups of hospitals classi fied by proportion of Medicare admissions. As in Figure 5 3, the average rate of HAI decreased for all three groups of hospitals although the rate of decrease was higher fo r G roups 1 and 3. However, analyzing the data by years as shown in Figure 5 4, the d ecrease was larger between 2007 and 2008 than between 2008 and 2009 Figure 5 5 shows that there was a reduction in the rate of hospital acquired conditions for all three categories of hospital groups of Medicare patient admissions. In this case, the reduc tio n was similar for groups 1 and 2 but smaller for group 2 indicating that hospitals with Medicare admissions above 50% had the smallest reduction in hospital acquired conditions post policy. Finally, Figure 5 6 shows that c ontrary to the pattern with th e hospital acquired infections, there was an increase in the rate of hospital acquired conditions between 2007 and 2008 followed by a sharp decrease between 2008 and 2009. Because Medicare announced the eight conditions that they will not pay for about a year before implementing the policy, hospitals might have started preparing or responding before the implementation date. To investigate this possibility, the three groups with varying proportion of Medicare patient admission were analyzed by the quarter i n which an admission took place Trend 1 depicts mean infection rates for admission in the first quarter of 2007, Trend 8 (last quarter of 2008) is when the policy was implemented, and Trend 12 is the last quarter of 2009.

PAGE 75

75 As shown in F igure 5 7, there does not appear to a pattern in the trend analysis for CAUTI except for G roup 3 (hospitals with more than 50% of admissions as Medicare patients ) The story is differen t for HAI. As shown in Figure 5 8, there was a consistent reduction in HAI per quarter and t he consistency is strongest for hospitals with more than 50% of Medicare admissions. As shown by Figure 5 9, HAC rates increased in the first 4 quarters and decreased consistently from quarter 5 for all groups of hospitals. These show that there is no evid ence of hospitals responding adequately to the policy in 2007 but they have responded positively since the first quarter of 2008. Multivariate Analysis This section analyzes the results of the HLM regressions for the main effects and full model effects fo r the three models of interest (CAUTI, HAI, and HAC). Table 5 4 summarizes the main effects of the log odds of both dummy variables and continuous variables. Because the dependent variable is a logit, the coefficients are in log odds and the main effects m ake it easier to interpret the interaction terms in the full model. Table 5 5 summarizes the full model effects of the log odds which includes interaction terms. This full model is a better fit than the main effect justifying the inclusion of interaction t erms Finally, Table 5 6 summarizes the odd ratios of the full model which is a transformation of the log odds from Table 5 5. Odds ratios are easier to interpret than log odds. The coefficients of s ome of the variables are consistently negative across all three models which mean they are predicted to reduce the log odds that a patient admitted with any of the corresponding infection or condition, acquired it in the hospital after 48 hours of being admitted. In contrast, a positive coefficient denotes an i ncrease in the log odds that the acquired condition was not present on admission.

PAGE 76

76 Independent Variables The results in Table 5 4 shows that patients had a lower probability of CAUTI post policy, patients in hospitals with higher proportion of Medicare admi ssions had higher probability of CAUTI, and patients in hospitals with higher previous year operating margins had lower probability of CAUTI. Similarly, patients were less likely to acquire HAI post policy than pre policy, patients in hospitals with higher Medicare admissions were more likely to acquire HAI, HAI. Finally, patients had lower probabilities of acquiring HAC post policy and lower probabilities in acq uiring HAC both in hospitals with higher proportion of Medicare admission or higher previous yea 4 and 5 5 describe the direction, and the specific effects will be discussed us ing the odds ratio from Table 5 6. Other Covaria tes the CAUTI M odel Age is positively correlated with CAUTI, meaning older patients are more likely to acquire CAUTI. Also, patients on Medicare HMO, commercial insurance or other government/self pay were more likely to get CAUTI than patients on Medicare. On the other hand, Medicaid patients were less likely than Medicare patients to acquire CAUTI. Blacks were less likely than Whites to acquire CAUTI and males were less likely to acquire CAUTI than females. The longer a patient stays on admission the more likely they acquire CAUTI. Patients who used the ICU were more prone to acquire CAUTI than those who did not. Patients with more co morbidities also had higher likelihood of acquiring CAUTI. Finally, patients at government owned hospitals were more likely than non government hospitals to acquire CAUTI, while patients in hospitals where more surgeries took place were more likely to acquire CAUTI.

PAGE 77

77 HAI Model Patients post policy were less likely to acquire HAI than patients pre policy. Mediastinitis is a rare condition and was not found in the data, so invariably HAI consisted of CAUTI and vascular catheter associated infections (VCAI). The significant decrease observed could be attributed mainly to VCAI. Patients in hospitals with higher proportion of Medicare admissions had a higher likelihood to obtain HAI. Similar to the CAUTI model, differences existed between patients depending on the type of insurance they had. Medicare patients did better than Medicare HMO, commercial, and other insurers apart from Medic aid. The more co morbidities you have the higher the likelihood to acquire HAI. Females and whites were more likely than males and blacks respectively to acquire HAI. Using the ICU or staying longer on admission was also associated with higher likelihood to acquire HAI. Volume of surgeries and proportion of surgeries per hospital had a negative correlation with acquiring HAI. This confirms the role of experience in positive health outcomes. HAC Model There was a reduction in the probability of acquiring a ny of the 8 targeted conditions (HAC) post policy. Patients in hospitals with better previous year operating margins were less likely to acquire HAC. When you put all conditions together, younger people were less likely to obtain HAC than older ones. Simi lar to the HAI and CAUTI models, patients with Medicare were less likely to acquire HAC than patients with Medicare HMO, commercial, or other insurances except Medicaid. Hispanics and Blacks were also less likely than whites to acquire any of the hospital acquired conditions. Being in an ICU or spending one more day on admission increases the likelihood of HAC. Unlike the HAI model, patients in hospitals with higher volumes of surgeries also had a higher probability of HAC. Patients in government owned hosp itals were more likely to obtain HAC than those in not for profit hospitals.

PAGE 78

78 Table 5 5 is the full model which includes interaction terms; POMCARE (which is interaction between policy and proportion of Medicare) and POOPMPA (an interaction between policy a nd operating margin per admission of the previous year). Patients in hospitals with higher previous year operating margins had higher likelihood of acquiring HAC post policy than pre policy. Patients in hospitals with higher proportion of Medicare admissio ns had higher likelihood of acquiring CAUTI, HAI, and HAC post policy than pre policy. This is contrary to the hypothesis that hospitals with higher proportion of Medicare admissions will have a higher incentive to decrease the likelihood of acquiring hosp ital infections or conditions post policy. Table 5 6 presents odds ratios for the full patient outcome model. Th e results presented in Tables 5 4 and 5 5 are in log odds and their interpretations are less intuitive. These logs are therefore exponentiated t o obtain the odd ratios. The results will be discussed based on their statistical significance starting with the independent variables of interest followed by patient characteristics and hospital characteristics. The odd of an event is the ratio of the pro bability of the event occurring to the probability of the event not occurring. For example, if the probability that it will rain is forecasted to be 60%, then the probability that it will not rain is 40% and the odds of it raining is 60/40=1.5, meaning it is 1.5 times more likely to rain than not to rain. The odds ratio on the other hand compares the odds of two events. For example, you can compare the odds of a female visiting the doctor with the odds of a male visiting a doctor. If the probability that a female will visit the doctor is 60% and the probability of a male visiting a doctor is 20% then the odds of a female visiting a doctor is 60/40 (1.5) and the odds for a male will be 20/80 (0.25). The odds ratio of females compared to males therefore is 1.5/0.25 (6). So the odds of a female visiting the doctor are 6 times the odds of a male. However, females are 3 times more

PAGE 79

79 likely to visit a doctor than males (60% to 20%) Odd of 1 means the likelihood of raining is the same as the likelihood of not rain ing (50 50 chance) but odds ratio of 1 means the odds of females and males are the same or the likelihood of a female visiting a doctor is the same as the likelihood of a male. Odd ratios can be converted to probabilities (odds/1+odds). Independent Variabl es The main independent variable of interest is POLICY. According to Table 5 6, the odd of a patient with CAUTI post policy is 0.93627 times the odds pre policy for hospitals with the average proportion of Medicare admissions (centered MCARE=0 ) and with the average previous operating margin (centered OPMPA=0). The inverse is 1.068068 which is the odds pre policy. Meaning the odds of a patient with CAUTI acquiring it in one of such hospitals pre policy is 1.068 times the odds after the polic y. In the main effects (Table 5 4) the odds ratio for Policy is 0.959 which in terms of probabilities means that post policy the probability of a patient with CAUTI acquiring it in the hospital is 0.49 compared with 0.51 pre policy. So you are less likely to acquire CAUT I post policy than pre policy. However, the coefficient is not statistically significant. For HAI, the odds post policy is 0.66805 times the odds pre policy for a hospital with typical Medicare admissions and typical previous year operating margins. In te rms of probabilities, the probability of a patient acquiring HAI post policy is 0.40 compared to 0.60 pre policy. In terms of odds ratio, the odds of a patient pre policy to acquire HAI is 1.5 times (1/0.66805) the odds of acquiring it post policy and this is statistically significant. Finally for HAC, the odds post policy is 0.94996 times the odds pre policy for a hospital with typical Medicare admissions and typical previous year operating margins. In terms of probabilities, the probability of a patient with one of the hospital acquired condition acquiring it in the hospital post policy is 0.49 compared with 0.51 pre policy (this is statistically significant).

PAGE 80

80 POOPMPA, an interaction term, seeks to estimate the impact of the policy on hospitals with incre asing operating margins. The odds ratio for HAC is 1.00001 which is the multiplicative factor for a unit increase in operating margin for post policy compared to pre policy. Because a unit increase is a dollar, the odds as currently measured are difficult to interpret. However, the estimate from Table 5 4 (0.00000074) can be multiplied by a 1000 and the exponent will represent the effect of an increase in $1000.00 of operating margin. When you do that, the odd ratio is 1.00074 meaning that a $1000 increase in operating margin increases the odds from pre to post policy by 1.00074. In terms of probabilities, an additional gain of $1000 in operating margins increases the probability of HAC post policy by about 0.07% compared to pre policy rates. Interestingly hospitals with higher OPMPA (previous year operating margin per admission) had lower likelihood of HAC but post policy their likelihood increased. This implies that successful hospitals are maintaining their stronger financial performance but not through reduction of HAC as predicted. POMCARE, the interaction term comparing the effect of the policy on the proportion of Medicare admissions per hospital in the previous year, shows an increase in the odds of acquiring CAUTI, HAI, and HAC post policy compared to pre policy. This is contrary to expectations but consistent with the ANOVA analysis. This is driven mainly by hospitals with higher than 50% of Medicare admission. Patient c haracteristics A one year increase in age increases the odds of acquiring CAUTI by 1.00649 and HAI by 1.02284, however it decreases the odds of acquiring HAC by 0.99834. This implies that the odds of acquiring a hospital acquired infection increases with age but the odds of acquiring any one of

PAGE 81

81 the 8 conditions (for example air embol ism, leaving foreign object in a patient after surgery, or blood incompatibilities) decreases with age. The odds of Medicare HMO patients in the sample acquiring CAUTI are 1.525 times the odds of Medicare patients. Similarly, the odds of Medicare HMO patie nts acquiring HAI are 1.41 times the odds of Medicare patients, and for acquiring HAC patients, 1.246 times the odds of Medicare patients. The odds of CAUTI patients with commercial insurance are about 2.5 times the odds of Medicare CAUTI patients, the odd s of HAI patients with commercial insurance are 1.8 times the odds of Medicare patients, and HAC patients with commercial insurance are 1.6 times the odds of HAC Medicare patients. The pattern is similar for patients without insurance or with other govern ment insurances except Medicaid 1.7 times the odds to acquire CAUTI, 1.8 times the odds to acquire HAI, and 1.4 times the odds to acquire HAC than Medicare patients. And these are all significant at the 1% level. An additional co morbidity increase the odd s of a patient acquiring CAUTI by 1.0112, HAI by 1.038, and HAC by 1.038.Females are more likely than men to acquire any of the conditions. The odds of a male acquiring CAUTI is 0.42262 the odds of females. In terms of probabilities, the probability of mal es is 0.3 compared to 0.7 of females. Alternatively, females are 2.4 times the odds of males to acquire CAUTI, about 1.9 times the odds of males to acquire HAI, and about 1.2 times the odds of males to acquire HAC. Similarly, whites are more likely than bl acks to acquire any of the three conditions of interest. The odds of whites are about1.8 times the odds of blacks to acquire CAUTI, 1.8 times the odds of acquiring HAI, and 1.5 times the odds of acquiring HAC. Patients who used the ICU have about 2 times the odds of getting CAUTI, 1.8 times the odds of getting HAI, and about 1.6 times the odds of getting an HAC than those who did not use

PAGE 82

82 the ICU. An additional day spent at the hospital increases the odds of a patient with CAUTI to have acquired it at the h ospital by 1.084, increases the odds of patients with HAI by 1.07, and increases the odds of patients with HAC by 1.07. Hospital c haracteristics Almost all the coefficients for hospital characteristics were not statistically significant. Surgery, a variabl e indicating the proportion of admissions due to surgeries was significant at the 5% level for HAI, with a 1% increase in the proportion of surgery admission leading to 0.98153 odds of a patient with HAI acquiring it while in the hospital. Thus, hospitals that do approximately more surgeries are less likely to have patients acquire HAI (at the 5% level of confidence). However for surgery as a volume measure (TOTSUR), an additional surgery performed by a hospital increases the odds of a patient acquiring all three conditions significantly at the 1% level of significance; these odds of acquiring these conditions due to an additional surgery increases by 1.00001 for patients with CAUTI, 100003 for patients with HAI, and by 1.00003 for patients with HAC. The res t of the hospital variables are not statistically significant although patients in government owned hospitals are more likely to acquire HAI or HAC at the 10% level of significance. Intercept To make interpretation of the intercept easier, the continuous v ariables were centered using the grand means. The intercept therefore represents the typical 70 year old white female (with CAUTI, HAI, or HAC) with Medicare who was not admitted to the ICU but admitted to an urban, non teaching, not for profit hospital, p re policy. For such a patient the odds that the CAUTI/HAI/HAC was not present at admission pre policy are 0.22716 for CAUTI, 0.13800 for HAI, and 0.09910 for HAC. The predicted probability of such a patient can be calculated using the formula (exp (coeffic ient)/1+exp (coefficie nt). From Table 5 5, the coefficient for the

PAGE 83

83 intercepts for CAUTI, HAI, and HAC are 1.4821, 1.9805, and 2.3116 respectively. Using the formula above, the predicted probability of such a patient before the policy corresponds to 0.18 5 for CAUTI, 0.121 for HAI, and 0.090 for HAC. To find the predicted probability after the policy, you add the coeff icients for policy from Table 5 5 ( 0.06585 for CAUTI, 0.4034 for HAI, and 0.05133 for HAC) to the coefficients of their intercepts ( 1.48 21, 1.9805, 2.3116 respectively) and use the formula above. The predicted probability of such a patient after the policy corresponds to 0.175 for CAUTI, 0.084 for HAI, and 0.086 for HAC). The impact of the policy on such a patient will be a reduction in the probability of acquiring CAUTI from 0.185 to 0.175, a reduction in the probability of acquiring HAI from 0.121 to 0.084, and a reduction in the probability of acquiring HAC from 0.090 to 0.086. Sensitivity Analysis To mimic the ANOVA analyses in which hospitals were categorized into 3 groups based on the proportion of Medicare admissions (those with less than 25% of admissions as Group 1, between 25 to 50% as Group 2, and above 50% as G roup 3), a sensitivity analysis was performed by replacing the cont inuous variable MCARE with the categorized variables (Mcare1, Mcare2, and Mcare3 as reference group). Additionally, age was categorized into children (less than 12 years), youth (between 12 and 35 years), adults (between 35 and 65 years) and elderly (above 65 years) to understand the impact of categorization of age on the models. Since the policy is targeted at the elderly, it is possible age could influence the outcomes differently depending on how it is modeled. Table 5 7 summarizes the full model versio n of the odds ratios of the categorized model. As shown in Table 5 7 categorization of Medicare admissions as well as age had a significant impact on the results. First of all the sta tistically significant decrease in hospital acquired conditions (HAC) in the full model of the continuous version ceases to be significant in this

PAGE 84

84 categorized version but they are both significant in their main effects versions as shown in the partial results of the categorized model of HAC These results still show that th ere is a decrease in HAC post policy with an odds ratio of 0.93497 instead of the 0.94137 in the continuous model. This means that the significant decrease shown earlier can be explained by the modifiers in the full model. Group 1 (Mcare1) hospitals (less than 25% of Medicare admissions) are 1.10189 times more likely than G roup 3 (more than 50% of Medica re admissions) to have HAC and G roup 2 hospitals are 0.89163 times the odds of the group 3 hospitals to acquire HAC. These confirm the negative correlation between Medicare admissions and HAC observed earlier with the continuous model. Medicare group 1 in the post policy era (POMCARE1) is 0.873 times the odds of group 3 in the pre policy era at 5% level of significance. Meaning the expected reduction in HAC post policy came from hospitals with lower proportion of Medicare admissions. When age is categorized into children, youth, adults and elderly instead of as a continuous variable, the results show that the odds of children under the age of 12 acquiring HAC are 1.8 times the odds of the elderly (above 65 years), 8 times the odds of the elderly to acquire CAUTI, and 0.18 times the odds of the elderly in acquiring HAI. In summary, there was a general reduction in hospital acquired infections and conditions pos t policy but the reduction came more from hospitals with less than 50% of Medicare admissions than those with more than 50% Medicare admissions. Hospitals with more previous year operating margins also had lower hospital acquired conditions. However, patie nts in those hospitals had lower probability of acquiring HAC post policy when you categorize Medicare and age but in the opposite direction when you model it as continuous variables. Additionally, the gender, age, type of insurance, the health status, len gth of stay in a hospital, and whether a

PAGE 85

85 patient was in the ICU influences the probability of HAI and HAC in an acute care hospital in Florida. Part 2 : Financial Performance of Acute Care H ospitals Descriptive Statistics This section focuses on private, short term acute care hospitals. Only hospitals with data in all three years were used. Out of the 248 hospitals, the final sample used for the analysis was 118 hospitals per year. Table 5 8 describes the variables and summarizes their means. All dollar va lues were adjusted for inflation using estimates from the Bureau of Labor and Statistics and 2007 as the reference year; 2008 dollars were adjusted by a factor of 1.038 and 2009 dollars were adjusted by a factor of 1.035. After adjusting for inflation rev enue, cost, and profits increased over the period 2007 to 2009. The typical hospital had about $200,000,000 net revenue, which includes both inpatient and outpatient revenues after discounts and deductions. Because the data set reported only total deductio ns and discounts inpatient revenue was calculated by inpatient charges less total discounts adjusted total discounts adjusted by the ratio of inpatient and outpatient revenues with the typical hospital having about $134,000,000 in inpatient revenues. Cos t represents total operating expenses, which includes both direct and indirect patient care expenses and the typical hospital incurred about a $196,000,000 in operating expenses. To analyze the impact of the policy on direct patient care expenses (DICOST), indirect patient care cost like support staff salaries were excluded. The typical hospital incurred about $112,000,000 in direct patient care expenses. Operating profit increased from approximately $5 million in 2007 to about $6 million in 2008 and about $10.7 million in 2009. The rate of hospital acquired conditions (HAC) increased from 1.7 % in 2007 to 2.3% in 2008 and 2.4% in 2009. Of the total hospital acquired conditions, 13.3% were not present on

PAGE 86

86 admission (RHAC) in 2007, 17.2% were not present on ad mission in 2008, and 13.5% were not present on admission in 2009. Conditions present on admission are mainly readmissions. A typical hospital in the sample had about 15,000 admissions per year; about 50% of admissions were Medicare and 14% were Medicaid pa tients. Outpatient visits averaged about 53,000 per year. The typical hospital had over 200 beds with an occupancy rate of 57% and an average length of stay of about 4.8 days. Of the sample, 57% were investor owned hospitals, 43% not for profit hospitals, 5% teaching hospitals, and 25% had a transplant program. The average overall case mix was about 1.3. The average patient care salaries/wages per FTE (PPERS) was about $16,573 and the average support staff salary per bed (PSUPP) was about $53,000. The avera ge price estimated to have been charged per admission in a given MSA was about $13,000. Unemployment rose from an average of 4.7% in 2007 to 11.6% in 2009 in the MSAs in the sample. The per capita income for a typical MSA was about $40,000. The average num ber of physicians per 1000 population was about 3 with a population density of about 880 per square mile. The typical market is highly concentrated with a Herfindahl index of 0.2 meaning a highly competitive market. ANOVA Analysis The ANOVA analysis involv es comparison of estimates of the dependent variables of a given model at four different levels of the proportion of Medicare admissions, at four different levels of the proportion of hospital acquired conditions not paid for by Medicare, and the three yea rs in the study (2007, 2008, and 2009). The tables below give the summaries, and the graphs give a pictorial presentation of the findings of the analysis. Because the cost and revenue variables in the multivariate analysis were logged to improve their line arity, these values are

PAGE 87

87 logged for the ANOVA analysis as well. The dependent variables analyzed included; direct inpatient care revenues (LNDIR), direct inpatient care expenses (PCREX), net patient care revenues (LREV), net patient care expenses (LCOST), o perating profit (OPMARG), and operating margins as a ratio o f net revenue (MARG). Medicare G roup 1 include hospitals with less than 38% Medicare admissions, Medicare G roup 2 include hospitals with between 38 % and 50% Medicare admissions, G roup 3 has betwee n 50% and 62% Medicare admissions, and group 4 has over 62% Medicare admissions. Of the 4 groups labeled infection; group 1 includes hospitals with less than 7% of hospita l acquired conditions (RHAC), G roup 2 include hospitals w ith between 7% and 14% of H AC, G roup 3 include hospitals with between14% and 21% of RHA C, and G roup 4 hospitals have over 21% of hospital acquired conditions. The groupings were based on the means (RHAC 14.6%, MCARE 51%) and standard deviations (RHAC 7.8, MCARE 12.5) of the distribu tion. Direct i npatient care r evenues (LNDIR) The results of Figure 5 11 show that LNDIR does not differ significantly across the three years, but LNDIR differs at the 1% level of significance across the different groups of hospitals with different levels o f RHAC (infection). Hospitals with different RHAC levels do not differ in the amount of inpatient revenues received per year across the three years. However inpatient revenues of hospitals with different RHAC differ significantly across the three years de pending on their proportion of Medicare admissions. Figure 5 9 explains where the differences occur. There were no difference s in average inpatient revenues among the groups between 2007 and 2 008 but the 2 G roups with Medicare admissions above 50% (G roups 3 and 4) had lower inpatient revenues than G roups 1 and 2 (less than 50% Medicare admission) from 2008 to 2009, implying that after the policy, without controlling for other factors, hospitals with less than 50%

PAGE 88

8 8 Medicare admissions received less inpatient revenue s than hospitals with higher than 50% Medicare admissions. Direct inpatient care expenses PCEXP The partial resu lts of the ANOVA displayed in Figure 5 13 show a similar pattern as the results of the inpatient care revenues discussed earlier. Ther e is a difference in inpatient care expenses among hospitals with different levels of RHAC depending on the proportion of Medicare admissions they have. However, contrary to expectations, the chan ge from 2008 to 2009 was larger for G roup 2 members than Gro ups 3 and G roup 4. I t was my expectation that hospitals with higher proportion of Medicare admissions will invest more in inpatient care in 2009 than hospitals with lower proportions of Medicare admissions to improve infection control in order to curtail t he potential loss of revenue after the policy. Net patient care revenues (LREV) LREV includes both inpatient and outpatient revenues but the results do not change the direction o f the discussion. As shown in Figure 5 15 the results are similar to the resu lts of the inpatient care revenues. The results show differences in mean revenues among the different groups both in terms of RHAC and the proportion of Medicare admissions but do not vary across the years. Hospitals with lower proportions of Medicare admi ssions had higher revenues than hospitals with higher proportion of Medicare admissions. Net patient care expenses (LCOST) Total patient care expenses follow the same pattern as inpatient care expenses. As shown by Figure 5 18 hospitals with lower proport ion of Medicare admissions spent more post policy than hospitals with higher proportion of Medicare admissions. There seems to be no change in patient care expenditure across the years for hospitals with higher proportion of Medicare admissions.

PAGE 89

89 Implying t hat the policy did not change the total amount they spent on patient care post policy. It could be because hospitals with lower proportion of Medicare admissions co uld afford the added investment but not those with higher Medicare admissions. Or they found alternative ways of curtailing the impact of the policy apart from investing extra to improve patient care. Operating profit (OPMARGIN) Operating profit is the difference between the net operating revenue and the net operating expense. There is a signif icant difference in operating profits across the years and among hospitals with different levels of RHAC. However, the change in operating profits does not differ across the years based on the proportion of hospital acquired conditions (labeled infection a bove) in each group but the differences in operating margins across the years differ significantly based on their proportion of Medicare admissions and proportion of hospital acquired conditions. From Figure 5 13, all groups of Medicare hospitals increased their operating margin post policy. Hospitals with lower proportion of Medicare admissions made higher operating margins than hospitals with higher proportion of Medicare admissions. Profit margin (MARG) The results displayed above shows significant diff erences in profit margins (operating margin/net patient revenue) across years, and across different levels of hospital acquired conditions. But these differences are not significantly different statistically depending on proportion of RHAC or proportion of Medicare admissions per hosp ital. The F igure 5 14 further highlights these findings From F igure 5 14 all groups of hospitals increased their profit margin s post policy and the policy does not seem to have had a negative impact on their profit margins Al though hospitals with lower proportion of Medicare admissions spent more on inpatient care they still made profit

PAGE 90

90 M ultivariate A nalysis This section summarizes the methods used and the results and analysis of the regression. The ANOVA findings could be d ue to other factors apart from the policy, the proportion of Medicare admissions, or the proportion of hospital acquired conditions. The R square of between 0.10 and 0.24 across the 6 models discussed above confirms that the bivariate model explains less t han 25% of the variability in the dependent variables. OLS was chosen for the regression analysis because it was a good fit for the data. The following diagnostics were performed to check the fit between the data and the assumptions of OLS: checks for infl uential plots and normality plots, variance inflation factor for collinearity, and white test for heteroscedasticity. Skewness ranged from 0.12 to 0.46 and kurtosis ranged from 0.855 to 3.0 4. There was a kurtosis of over 6 due to an influential point which was eliminated. The most significant observation of deviation is the collinearity due to the interaction terms and the VIF for unemployment was also high about 20. Table 5 8 summarizes t he results of the regression analysis for all six models, with each variable representing the partial effect of the variable holding all the other variables constant. LDIREV is the log inpatient revenues, LREV is the log net operating revenue, LCOST is the log net operating expenses, LDICOST is the log of direct patient care expenses, MARG is the operating profit margin, and PROFIT is the operating profit. Because the revenues and cost variables were logged, the coefficients in the tables can be interpreted as the percent change in the dependent variable given a one unit change in the independent variable. RHAC is the proportion of hospital acquired conditions not present on admission. 5%, net operating revenue by 0.5%, and direct inpatient expenses by about 0.6%. A 1% percentage

PAGE 91

91 increase in Medicare admissions reduces both inpatient and net patient revenue, and inpatient expenses by about 0.5%. Covariates of operating margin and operati ng profit Financial years 2009 and 2008 operating profit and operating margins were higher than 2007. Several factors could account for the increa sed profitability; from Table 5 8 admissions were higher in 2009 than in 2008 and 2007, additionally hospitals spent less on patient care salaries and wages as well as support staff salaries in 2009 than in 2007 and 2008. This could be in anticipation of the loss of revenue from the Medicare policy. Price was a significant predictor of both operating profits and o perating margins. Prices were higher in 2009 than in 2008 and 2007 and this could be due to a mix of cases admitted in 2009 compared to 2008 and 2007. An additional admission increases operating profit by $1038. For profit hospitals made about $4 million i n operating margins and 2.8% more in operating margins than not for profit hospitals. Teaching hospitals made about $ 2.7 million less in operating profits and 4.5% less in operating margins than non teaching hospitals. Communities with higher per capita i ncome made less profit than those with lower per capita income. A percentage increase in occupancy rate increases profit margin by about 0.16 and bigger hospitals made less operating profits than smaller hospitals. Covariates of net revenue and inpatient r evenue A 1% increase in price increases net revenue by about 0.003%. As expected, an additional admission increases both inpatient revenue and net revenue. Unexpectedly, an additional increase in average length of stay decreases net revenue. For profit ho spitals made less net revenue than not for profit hospitals and a higher case mix increased both net revenue and inpatient revenue. A 1% increase in Medicaid patients increased inpatient revenue by about 0.9%. As expected higher occupancy rate was associat ed

PAGE 92

92 with higher net revenue and inpatient revenues. Finally, bigger hospitals made more inpatient and outpatient revenues than smaller hospitals. Covariates of inpatient expenses and total operating expenses As expected admissions increase patient care exp enses but each additional day of length of stay decreased both inpatient expenses and total operating expenses by about 6%. Hospitals with sicker patients (CASEMIX) spent more on total operating and inpatient expenses. Hospitals with a transplant program s pent more both patient expenses and total operating expenses than hospitals without a transplant program. Hospitals in MSAs with higher unemployment rates spent less on both patient care and total operating expenses. Occupancy rates and bed size are associ ated positively with both patient care and total patient expenses. Finally, the higher the number of full time equivalent patient care staff, the higher the operating and patient care expenses. Intercept Because the continuous variables are centered, the intercepts represent a typical non teaching, not for profit hospital with less than 100 beds and without a transplant program in 2007. Such a hospital received about $45,677,977 in inpatient revenue, $75, 637,725 in net operating revenue, $76, 309,125 in o perating cost, out of which $41,595,982 were for direct patient care expenses. These hospitals also made about 0.86% loss and an operating margin of $12,134,009 in 2007. Full model analysis The goal of this section is to further explore the impact of the p olicy on the financial performance of hospitals in the sample by examining how the impact is modified by the proportion of Medicare admissions and prop ortion of RHAC per hospital. Table 5 9 summarizes the full model which includes the interaction terms. MC AREYR08 and MCAREYR09 are the interactions of Medicare admission in 2008 and 2009 respectively

PAGE 93

93 compared with 2007. Similarly, RHACYR08 and RHACYR09 compare hospitals in 2008 and 2009 with those in 2007 based on the proportion of hospital acquired condition s. Finally, MCRHAYR08 and MCRHAYR09 compares 2008 and 2009 with 2007 based on the proportion of hospital acquired conditions and proportion of Medicare admissions. As shown in Table 5 9, almost all the interaction terms were not statistically significant w ith the exception of inpatient care of MCAREYR08 and operating profit of MCRHAYR08. This means that ho spitals with higher proportion of Medicare admissions spent more on inpatient care in 2009 than in 2008 but patient with higher proportion of Medicare adm issions and higher proportion of hospital acquired conditions made more operating profit in 2008 than in 2009. Sensitivity Analysis Three different sensitivity analyses were performed on the financial performance model. The first one looked at all hospita l acquired conditions whether they were present on admission or not. This was to assess the impact of DRG coding problems or abuse of the present on admi ssion indicator. As shown in Table 5 10, the patterns are similar for the covariates; the major differ ences are the hospital acquired conditions. When you compare hospitals based on HAC (all conditions present on admission or not) instead of RHAC (conditions not present on admission), revenues for hospitals with RHAC increase with an increase in rate of RH AC but percentage increases in HAC do not affect HAC hospitals. Secondly whereas total patient care expenditure increases in RHAC it rather decreases in HAC. This implies that hospitals with higher hospital acquired conditions spend less on patient care an d the rate of hospital acquired conditions does not affect its revenues significantly. The ANOVA analysis discussed earlier categorized hospitals into 4 groups according to the proportion of Medicare admissions as well as the proportion of hospital acquir ed conditions. The next sensitivity analysis looked at the impact of using categorical versions of these variables

PAGE 94

94 in the regression m odels. As shown in Tables 5 11 and 5 1 2, there were no significant differences in the impact of categorization on the tren d of results. Similarly, hospitals with higher proportion of Medicare admissions and higher proportion of hospital acquired conditions received less revenue post policy. The third sensitivity analysis was to use a mixed model instead of OLS. The rationale was to assess if the clustering due to time imposed by the panel nature of the study design influences the results. In the mixed model each subject (hospital) has three observations (for 2007, 2008, and 2009) which understandably are correlated. Mixed mode ls handle correlated errors better than OLS and models the intercepts of each hospital as an outcome in the second level model. The results again were comparable to the OLS in that the interaction terms were not significant but the surprising difference be tween the two methods is that RHAC effect on revenue and cost were not statistically significant in the Mixed Model but were significant in the OLS Model. Meaning infections rates were not a significant predictor of revenue and cost in the mixed model.

PAGE 95

95 Ta ble 5 1 Means of Level 1 Variables p re p o ost p Variable Description CAUTI HAI HAC Dependent Variables Pre (5,489) Post (4,848) Pre (7,392) Post (9,234) Pre (72,998) Post (66,060) RCAUTI Percentage o f all patients with CAUTI whose condition were not present on admission (NPOA) 19% 18% RHAI Percentage with hospital acquired infections NPOA 15% 10% RHAC Percentage with hospital associated conditions NPOA 16% 15% Independent Variables AGE Age at admission 71.60 70.52 66.19 61.96 70.79 69.95 MEDICARE Percentage of patients with Medicare as primary source of payment 72% 70% 63% 59% 65% 63% MCAREHMO Percentage with Medicare HMO 10% 11% 10% 10% 12% 13% MEDICAID Percentage with Medicaid as primary source of payment 5% 6% 7% 10% 7% 7% MCAIDHMO Percentage with Medicaid HMO 2% 2% 3% 3% 2% 3% COMMERCIAL Percentage with Private insurance 8% 8% 13% 14% 9% 10% OTHINS Percentage with other insurance 4% 4% 5% 5% 5% 5% COMORB Number of co morb idities 13.70 14.56 12.97 13.09 13.75 14.93 MALE Male coded=1 0.60 0.59 0.57 0.54 0.50 0.50 WHITE Proportion of whites 0.77 0.76 0.74 0.69 0.68 0.66

PAGE 96

96 Table 5 1 Continued Variable Description CAUT I HAI HAC BLACK Proportion of blacks 0.14 0.16 0.1 6 0.21 0.18 0.19 HISPANIC Proportion of Hispanics 0.06 0.05 0.07 0.07 0.10 0.10 OTHRACE Proportion of other races 0.03 0.02 0.03 0.03 0.04 0.05 ICU Proportion of patients who used the ICU 0.39 0.40 0.38 0.39 0.45 0.48 LOSDAYS Days spent before dischar ged 10.16 9.59 10.03 9.48 12.01 10.97 Table 5 2 Means of Level 2 V ariables per year; 2007, 2008, 2009 Variable Description 2007 (n=185) 2008 (n=177) 2009 (n=179) MCARE admissions 0.50 0.49 0.51 OPMPA Previou admission ($) 35 90 98 PROCESS Average rate of surgical care improvement reported on Medicare hospital compare website 94.50 94.45 94.48 SURGERY Prop. of admissions that are surgery 0.29 0.30 0.30 CATH Prop. of admissions with catheterization 0.41 0.41 0.41 PROFESSIONAL Presence of infection control specialist 0.68 0.69 0.69 BIG Bed size above 200 0.37 0.38 0.39 MODERATE Bed size between 100 and 200 0.46 0.47 0.47 SMALL Bed size below 100 0.17 0.15 0.14 TEACH Prop. of teaching hospitals 0.03 0.03 0.03 MEDDIR Presence of a medical director 0.77 0.77 0.77 NURSEFTE RN & LPN FTEs/inpatient days 0.36 0.36 0.36 RNFTE RN FTEs/inpatient days 0.32 0.32 0.32 RURAL No. of rural hospitals 0.15 0.13 0.12

PAGE 97

97 Table 5 3. Bivariate od ds ratio of dependent variables by policy CAUTI HAI HAC Time Odds Confidence Interval Odds Confidence Interval Odds Confidence Interval Pre policy 1.0716 0.9883 1.1619 1.5074 1.3868 1.6386 1.0828 1.0563 1.1100 Post policy 0.9843 0.9663 1.0026 0.9459 0 .9350 0.9568 0.9857 0.9813 0.9902 Odds Ratio 0.9185 0.8317 1.0145 0.6275 0.5708 0.6898 0.9104 0.8841 0.9374 Figure 5 1 CAUTI for MEDICARE Groups (1, 2 & 3) vs. POLICY (0, 1)

PAGE 98

98 Figure 5 2 CAUTI for M EDICARE Group 1 (<25%), 2 (25 50%), & 3 (>50%) vs. YEAR (2007 to 2009) Figure 5 3 HAI for MEDICARE Groups 1, 2, &3 vs POLICY (0,1)

PAGE 99

99 Figure 5 4 HAI for MEDICARE Groups 1 (<25%), 2 (25 50%), & 3 (>50%) vs. YEAR (2007 to 2009) Figure 5 5. HAC for MEDICAR E Groups 1 (<25%), 2 (25 50%), & 3 (>50%) vs. POLICY

PAGE 100

100 Figure 5 6 HAC for MEDICARE Group 1 (<25%), 2 (25 50%), & 3 (>50%) vs. YEAR (2007 to 2009)

PAGE 101

101 Figure 5 7. CAUTI for MEDICARE Groups 1 ( <25%), 2 (25 50%), & 3 (>50%) vs TR END

PAGE 102

102 Figure 5 8. HAI for MEDICARE Groups 1 (<25%), 2 (25 50%), & 3 (>50%) vs. TREND

PAGE 103

103 Figure 5 9. HAC for MEDICARE Groups 1 (<25%), 2(25 50%), & 3 (>50%) vs. TREND

PAGE 104

104 Table 5 4. GHLM Regression r esults of main e ffects of patient outcome variables Varia ble Description CAUTI HAI HAC Independent variables POLICY Period after policy 0.04176 0.4159*** 0.06042*** MCARE Proportion of admissions with Medicare 0.008676 0.01495*** 0.00582* OPMPA Previous year operating margin per admission 0.00010* 0.0004 0.00001*** Patient characteristics AGE Age at admission 0.006628*** 0.02286*** 0.00165*** MCAREHMO Medicare HMO 0.4114*** 0.3421*** 0.2203*** MEDICAID Medicaid as primary source of payment 0.3148* 0.09856 0.06774* MCAIDHMO Medi caid HMO 0.03053 0.07735 0.09097 COMMERCIAL Private insurance 0.9344*** 0.5754*** 0.4976*** OTHINS Other insurance 0.5534*** 0.6235*** 0.3443*** COMORB Number of co morbidities 0.01135* 0.03679*** 0.03808*** MALE Male coded=1 0.8231*** 0 .6102*** 0.18967*** BLACK Number of blacks 0.4855*** 0.5620*** 0.4055** OTHRACE Other races 0.09582 0.04080 0.02949 ICU Number of patients at ICU 0.7000*** 0.5824*** 0.4893*** LOSDAYS Days spent before discharged 0.08067*** 0.066 85*** 0.06853*** Hospital characteristics SURGERY Proportion of surgeries/hosp. 0.00891 0.01758** 0.009595** TOTSUR Total number of surgeries/hosp 0.000015*** 0.000027*** 0.000023*** CATH Proportion of catheterization 0.00023 0.000342 0 .000220

PAGE 105

105 Table 5 4. Continued Variable Description CAUTI HAI HAC Presence of infection control specialist 0.02792 0.1463 0.1765 Variable Description CAUTI HAI HAC SMALL < 100 beds 0.1459 0.1909 0.05092 MODERATE 100 200 beds 0.17 90 0.2148 0.1454 TEACH Teaching hosp. 0.1347 0.1190 0.3113 F0R PROFIT Investor owned 0.09271 0.1839 0.1297 GOVT government 0.3329* 0.4148* 0.2605* MEDDIR Presence of Med. director 0.1133 0.1730 0.1374 NURSEFTE LPNs & RNs FTE 0.01898 0.03671 0.01928 RNFTE Only RN FTE 0.02337 0.04042 0.01103 RURAL Rural hospital 0.7244 0.4400 0.1491 INTERCEPT 1.4403*** 1.9801*** 2.3107*** Key: *** means 1% significance, ** means 5% significance and means at 10% significance Table 5 5 GHLM Regression results of full m odel for patient outcome variables Variable Description CAUTI HAI HAC POLICY Period post policy 0.06585 0.4034*** 0.05133*** POOPMPA POLICY*OPMPA. 0.0000019 0.000013 0.00000074*** POMCARE MCARE post policy 0.01194** 0.01362*** 0.003699*** MCARE Medicare admissions 0.001003 0.009081 0.000686** OPMPA per admission 0.00010* 0.00004*** 0.00002*** Patient characteristics AGE Age at admiss ion 0.006473*** 0.02258*** 0.00167*** MCAREHMO Medicare HMO 0.4224*** 0.3439*** 0.2198*** MEDICAID Medicaid as primary source of payment 0.3221* 0.09770 0.06790*

PAGE 106

106 Table 5 5. Continued Variable Description CAUTI HAI HAC COMMERCIAL Private insura nce 0.9138*** 0.5774*** 0.4975*** OTHINS Other insurance 0.5519*** 0.6098*** 0.3423*** COMORB Number of co morbidities 0.01114* 0.03707*** 0.03800*** MALE Male coded=1 0.8613*** 0.6150*** 0.1859*** BLACK Number of blacks 0.4538*** 0.56 38*** 0.4051*** HISPANIC Number of Hispanics 0.07223 0.06970 0.06792** OTHRACE Other races 0.09533 0.05114 0.02736 ICU Number of patients at ICU 0.7450*** 0.5823*** 0.4899*** LOSDAYS Days spent before discharged 0.08065*** 0.06 671*** 0.06858*** Hospital characteristics SURGERY Proportion of surgeries/hosp. 0.00619 0.01864** 0.008626* TOTSUR Total number of surg./hosp. 0.00000761*** 0.000028*** 0.00003 *** CATH Proportion of catheterization 0.00013 0.000382 0.000240 Presence of infection control specialist 0.004677 0.1608 0.1777 SMALL < 100 beds 0.1335 0.1725 0.04689 MODERATE 100 200 beds 0.1972 0.2099 0.1482 TEACH Teaching hosp. 0.1459 0.1332 0.3155 F0R PROFIT Investor owned 0.07454 0.1806 0.1295 GOVT government 0.2713 0.4103* 0.2592* MEDDIR Presence of Med. director 0.1211 0.1819 0.1375 NURSEFTE LPNs & RNs FTE 0.01803 0.03666 0.01967 RNFTE Only RN FTE 0.02184 0.04047 0.01161 RURAL Rural hospital 0.7206 0.4332 0.1563 INTERCEPT 1.4821*** 1.9805*** 2.3116*** Key: *** means 1% significance, ** means 5% significance and means at 10% significance

PAGE 107

107 Table 5 6 O dds ratio for full model Variable Description CAUTI H AI HAC Independent variables POLICY Period post policy 0.93627 0.66805*** 0.94996*** POOPMPA POLICY*OPMPA 1.00000 1.00001 1.00001*** POMCARE POLICY*MCARE 1.01166** 1.01371*** 1.00371*** MCARE Proportion of Medicare admissions 1.000100 1.00912 0.9 9317** OPMPA margin per admission 0.9999* 0.99996 0.99998*** Patient characteristics AGE Age at admission 1.00649*** 1.02284*** 0.99834*** MCAREHMO Medicare HMO 1.52557*** 1.41043*** 1.24580*** MEDICAID Medicaid as primary source of payment 0.72460* 0.90692 0.93435* MCAIDHMO Medicaid HMO 0.95160 1.08487 0.91318 Commercial Private insurance 2.49377*** 1.78137*** 1.64455*** OTHINS Other insurance 1.73648*** 1.84003*** 1.40823*** COMORB Number of co morbidities 1.01120* 1. 03777*** 1.03873*** MALE Male coded=1 0.42262*** 0.54063*** 0.83038*** BLACK Number of blacks 0.63518*** 0.56903*** 0.66690*** OTHRACE Other races 1.10220 0.95015 1.02773 ICU Number of patients at ICU 2.10638*** 1.79008*** 1.63207*** LOSDAYS Days spent before discharged 1.08399*** 1.06898*** 1.07098*** Hospital characteristics SURGERY Proportion of surgeries/hosp. 0.99382 0.98153** 1.00866* TOTSUR Total number of surg./hosp. 1.00001*** 1.00003*** 1.00003*** CATH Proportion of catheteriz ation 0.99987 1.00038 1.00024 Presence of infection control specialist 1.00469 1.17446 1.19444 SMALL < 100 beds 0.87506 0.84155 0.95419 MODERATE 100 200 beds 0.82103 0.81067 1.15975 TEACH Teaching hosp. 1.1506 1.14245 1.37093 F0R P ROFIT Investor owned 0.92817 0.83479 1.13822

PAGE 108

108 Table 5 6 Continued Variable Description CAUTI HAI HAC GOVT government 1.31173 1.50731* 1.29588* MEDDIR Presence of Med. director 0.88595 0.83365 0.87154 NURSEFTE LPNs & RNs FTE 1.01820 1.03734 0.98052 RNFTE Only RN FTE 0.97839 0.96034 1.01168 RURAL Rural hospital 0.48646 0.64842 0.85759 INTERCEPT 0.22716*** 0.13800*** 0.09910*** Key: *** means 1% significance, ** means 5% significance and means at 10% significance Table 5 7. Odds ra tio for full model of the categorized age and Medicare admissions model Variable Description CAUTI HAI HAC Independent variables POLICY Period post policy 1.00918 0.70894*** 0.98268 POOPMPA POLICY*OPMPA 0.99995 0.99997 0.99997*** POMCARE1 POLICY*MC ARE 1 0.76331 0.65463 0.83335** POMCARE2 POLICY*MCARE2 0.87949 0.84014 0.097150 MCOPMPA1 Mcare1*OPMPA 0.99981 0.99992 0.99987* MCOPMPA2 Mcare2*OPMPA 1.00014 1.00000 1.00001 POMCPA1 Policy*MCARE1*OPMPA 1.00014 1.00010 1.00009** POMCPA2 Policy*MCARE2*OP MPA 1.00005 1.00006 0.99997 MCARE1 Prop. of Medicare admissions <25% 0.45735* 0.50647* 1.12142 MCARE2 Prop. of Medicare admissions 25 to 50% 0.62689*** 0.59789*** 0.86960** OPMPA margin per admission 0.99990 0.99998 1.00002 Pa tient characteristics CHILDREN Less than 12 years 8.03988*** 0.18202*** 1.79120*** YOUTH Between 12 and 35 0.50836*** 0.29493*** 1.10387** ADULT Between 35 and 65 0.78058*** 0.47406*** 0.86401*** MCAREHMO Medicare HMO 1.52255*** 1.39347*** 1. 24595*** MEDICAID Medicaid as primary source of payment 0.69306** 0.94783 0.97915

PAGE 109

109 Table 5 7. Continued Variable Description CAUTI HAI HAC COMMERCIAL Private insurance 2.65487*** 1.87083*** 1.77926*** OTHINS Other insurance 1.83255*** 1.96271*** 1.514 64*** COMORB Number of co morbidities 1.01093* 1.03739*** 1.03954*** MALE Male coded=1 0.43153*** 0.51183*** 0.83600*** BLACK Number of blacks 0.62343*** 0.57078*** 0.67301*** HISPANIC Number of Hispanics 1.05618 0.91318 0.89397*** OTHRACE Other races 1.08382 0.91586 0.99527 ICU Number of patients at ICU 2.08115*** 1.90819*** 1.63729*** LOSDAYS Days spent before discharged 1.08347*** 1.06767*** 1.07140*** Hospital characteristics SURGERY Proportion of surgeries/hosp. 0.98623* 0.98187** 1.00731 TOTSUR Total number of surgeries/hosp. 1.00002*** 1.00002*** 1.00003*** CATH Proportion of catheterization 0.99981 1.00036 1.00018 Presence of infection control specialist 1.07277 1.15077 1.23619 SMALL < 100 beds 0.88625 0.73874 0.96280 MODERA TE 100 200 beds 0.85153 0.82091 1.12926 TEACH Teaching hosp. 1.28475 1.26508 1.27755 F0R PROFIT Investor owned 0.94931 0.89231 1.09956 GOVT Government 1.39127 1.49730* 1.27201 MEDDIR Presence of Med. Director 0.89876 0.83028 0.88730 NURSEFTE LPNs & RNs FTE 0.99606 1.00759 0.97196 RNFTE Only RN FTE 1.00095 0.99162 1.02121 RURAL Rural hospital 0.55745 0.79481 0.82026 INTERCEPT 0.28867*** 0.23507*** 0.10436*** Key: *** means 1% significance, ** means 5% significance and means at 10% significa nce

PAGE 110

110 Table 5 8. Means of financial question variables per year; 2007, 2008, 2009 Variable Description 2007 (N=118) 2008 (N=118) 2009 (N=118) Dependent variables REV Total operating revenue 198,168,235 202,580,840 214,923,663 DIREV Inpatient revenue per hospital 130,387,154 131,568,585 137,261,951 COST Total operating expenses 193,267,539 196,634,903 204,395,549 DICOST Patient care expenses 108,816,314 111,296,229 115,087,862 PROFIT Operating margin 4,931,294 5,945,937 10,648,671 Independent variab les HAC proportion of admissions with hospital acquired conditions 1.7 2.3 2.4 RHAC Proportion of HAC not present on admission 13.3 17.2 13.5 MCARE Proportion of Medicare admissions 50.4 51.4 51.7 Other covariates BEDSIZE Indicator of the number of beds from 0 4 2.4 2.4 2.4 TEACH Proportion of teaching hospitals 0.05 0.05 0.05 FP Proportion of for profit hospital 0.57 0.57 0.57 CASEMIX Overall case mix 1.3 1.3 1.3 ALOS Average length of stay 4.8 4.8 4.8 ADM Number of admissions 15,303 15,344 15,810 OUTV Outpatient admissions 53,800 54,999 50,210 OCCURATE Occupancy rate (%) 58 57 56 TRANS Proportion of transplant hospitals 0.25 0.25 0.25 POPDEN Population density/square mile per MSA 870 870 892 MDPOP MDs per 1000 population per MSA 2.5 2.5 2.5 PPERS Ave. salary/ wages of pc workers per MSA 16,573 13,935 14,316 PSUPP Ave. salary/ wages of support staff per MSA 55,768 53,587 46,414 PRICE Average price per admission in MSA 12,663 12,885 13,247 UNEMP Unemployment rate in MSA 4.7 8.0 11.6 INCOME Per capita income per MSA 39,614 39,632 38,480 HHI Herfindahl Hirschman Index per MSA 0.2 0.2 0.2

PAGE 111

111 Effect Estimate Error DF t Value Pr > |t| Intercept 2.2952 0.1453 166 15.79 <.0001 policy 0.06724 0.01797 13E4 3.74 0.0002 mcare1 0.09696 0.1161 13E4 0.84 0.4035 mcare2 0.1147 0.06445 13E4 1.78 0.0753 children 0.5423 0.08885 13E4 6.10 <.0001 youth 0.07871 0.04737 13E4 1.66 0.0966 adult 0.1337 0.02408 13E4 5.55 <.0 001 Figure 5 10. Partial results from categorized model of HAC Dependent Variable: lnDir Sum of Source DF Squares Mean Square F Value Pr > F Model 46 72.3805866 1.5734910 2.88 <.0001 Error 307 167.7355309 0.5463698 Corrected Total 353 240.1161175 R Square Coeff Var Roo t MSE lnDir Mean 0.301440 4.028643 0.739168 18.34782 Source DF Type I SS Mean Square F Value Pr > F YEAR 2 0.07629872 0.03814 936 0.07 0.9326 Infection 3 29.80519224 9.93506408 18.18 <.0001 YEAR*infection 6 2.74413649 0.45735608 0.84 0.5420 YEAR*infecti*Medicar 35 39.7549 5914 1.13585598 2.08 0.0006

PAGE 112

112 Figure 5 11. Partial results of inpatient revenues Figure 5 12. Log Inpatient Revenue vs. Medicare year Infection rate

PAGE 113

113 Dependent Variable: pcexp Sum of Source DF Squares Mean Square F Value Pr > F Model 46 1.3396642E18 2.9123135E16 2.45 <.0001 Error 307 3.6486471E18 1.1884844E16 Corrected Total 353 4.9883113E18 R Square Coeff Var Root MSE pcexp Mean 0.268561 97.56936 109017631 111733468 Source DF Type I SS Mean Square F Value Pr > F YEAR 2 2.3544455E15 1.1772228E15 0.10 0.9057 infection 3 3.481704E17 1.160568E17 9.77 <.0001 YEAR*infection 6 5.9331692E16 9.8886154E15 0.83 0.5458 YEAR*infecti*Medicar 35 9.2980768E17 2.6565934E16 2.24 0.0002 Figure 5 13. Partial results of inpatient expenses

PAGE 114

114 Figure 5 14. Log Patient care expenses vs. Medica re year Infection rate

PAGE 115

115 Dependent Variable: lrev Sum of Source DF Squares Mean Square F Value Pr > F Model 46 70.2692094 1.5275915 3.17 <.0001 Error 307 148.1398171 0.4825401 Corrected Total 353 218.4090266 R Square Coeff Var Root MSE lrev Mean 0.321732 3.695915 0.694651 18.79511 Source DF Type I SS Mean Square F Value Pr > F YEAR 2 0.26986004 0.13493002 0.28 0.756 3 Infection 3 25.67009344 8.55669781 17.73 <.0001 YEAR*infection 6 2.58103284 0.43017214 0.89 0.5013 YEAR*infecti*Medicar 35 41.74822311 1.19280637 2.47 <.0001 Figure 5 15. Partial results of net patient revenue

PAGE 116

116 Figur e 5 16. Net Revenue vs Medicare Infection rate

PAGE 117

117 Dependent Variable: lcost Sum of Source DF Squares Mean Square F Value Pr > F Model 46 66.6621932 1.4491781 3.11 <.0001 Error 307 142.8809236 0.4654102 Corrected Total 3 53 209.5431168 R Square Coeff Var Root MSE lcost Mean 0.318131 3.634436 0.682210 18.77072 Source DF Type I SS Mean Square F Value Pr > F YEAR 2 0.07543851 0.03771925 0.08 0.9222 infection 3 23.94074367 7.98024789 17.15 <.0001 YEAR*infection 6 2.32800470 0.388 00078 0.83 0.5446 YEAR*infecti*Medicar 35 40.31800633 1.15194304 2.48 <.0001 Figure 5 17. Partial results of net cost

PAGE 118

118 Figure 5 18. LCOST vs. Medicare year RHAC Dependent Variable: opmarg Sum of Source DF Squares Mean Square F Value Pr > F Model 46 2.8971517E16 6.2981558E14 1.87 0.0011 Error 3 07 1.0354944E17 3.3729459E14 Corrected Total 353 1.3252095E17 R Square Coeff Var Root MSE opmarg Mean 0.218618 255.9555 18365582 7175301

PAGE 119

119 Sour ce DF Type I SS Mean Square F Value Pr > F YEAR 2 2.1961222E15 1.0980611E15 3.26 0.0399 infection 3 3.0066409E15 1.0022136E15 2.97 0.0321 YEAR*infection 6 1.3996356E15 2.3327259E14 0.69 0.6566 YEAR*infecti*Medicar 35 2.2369118E16 6.3911766E14 1.89 0.0024 Figure 5 19. Partial results of operating margins Figure 5 20. O perating profit v s. Medicare Year Infection rate

PAGE 120

120 Dependent Variable: marg Sum of Source DF Squares Mean Square F Value Pr > F Model 46 3570.21535 77.61338 1.12 0.2808 Error 307 21216.92925 69.11052 Corrected Total 353 24787.14460 R Square Coeff Var Roo t MSE marg Mean 0.144035 404.6281 8.313274 2.054547 Source DF Type I SS Mean Square F Value Pr > F YEAR 2 576.374936 288.187 468 4.17 0.0163 infection 3 507.680111 169.226704 2.45 0.0637 YEAR*infection 6 306.398134 51.066356 0.74 0.6187 YEAR*infecti*Medicar 35 2179.76 2172 62.278919 0.90 0.6326 Figure 5 21. Partial results of profit margin

PAGE 121

121 Figure 5 22. Profit Margin vs. Medicare year Infection rate

PAGE 122

122 Table 5 9 Resu l ts of m ain effects of RHAC (hospital acquired conditions not present on admission) on the financial performance of acute care hospitals Variable LDIREV LREV LCOST LDICOST MARG PROFIT RHAC 1.005203** 1.005033*** 1.0054 1.00566*** 0.0407 67157 MCARE 0.998112 0.994505*** 0.9947*** 0.99460*** 0.0114 168962 YR09 0.998311 1.030413 0. 996516 1.00095 2.7503*** 5487991*** YR08 0.976003 0.989476 0.982898 0.980532 0.4497 1071206** PRICE 1.00002 1.000031** 0.0009* 1765** ADMS 1.000021*** 1.000018*** 1.00002*** 1.00002*** 0.00006 1038*** ALOS 0.991338 0.943 801** 0.94742*** 0.9415*** 1.30414 681900 FP 1.01218 0.880007*** 0.86419*** 0.84358*** 2.79238*** 3977077** TEACH 0.99107 1.13325* 1.18813*** 1.16084** 4.52970* 27562960*** CASEMIX 2.73046*** 2.19319*** 2.30514*** 2.70718*** 1.29491 2487814 T RANS 1.031651 1.052923* 1.053133** 1.056932** 0.66339 2029984 HHI 0.889603 1.049234 1.065922 1.102422 4.69206* 2906479 MCAID 1.009505*** 1.003797 1.003998* 1.003536 0.03150 69231 INCOME 1.000003 1.000002 1.000001 1.000003 0.0002** 367** UNEMP 0.989684 0.979278 0.97412** 0.978925* 0.71202 483959 OCCRATE 1.013328*** 1.011415*** 1.00920*** 1.01072*** 0.15537*** 43235 BEDSIZE 1.333931*** 1.332424*** 1.32514*** 1.34362*** 0.08534 3587455*** POPDEN 0.999995 0.999984 0.999 99 0.999984 0.00001 8.9 MDPOP 0.995968 0.971999* 0.97927 0.975934* 0.37374 1953972* PPERS 1.000004*** 1.000004*** 0.00002 30.4 PSUPP 1.000000 1.000000 0.00005 17.1 INTERCEP T 45677977*** 75637726*** 76309126*** 41595982*** 0.8622 3 12134009*** Key: *** means 1% significance, ** means 5% significance and means at 10% significance

PAGE 123

123 Table 5 10 Results for f ull model effects of RHAC (hospital acquired conditions not present on admission) on financial performance of acute care hospitals Variable LDIREV LREV LCOST LDICOST MARG PROFIT RHAC 1.0036 1.004149 1.004038* 1.004339* 0.11167 69429 MCARE 0.997004 0.9938*** 0.994545*** 0.994117*** 0.02293 142378 YR09 0.998491 1.031094 0.998172 1.001201 2.71016*** 5595475*** YR08 0.9776 05 0.991715 0.984669 0.982554 0.54923 1307923 MCAREYR09 1.002844 1.002132 1.001031 1.002102 0.08248 32576 MCAREYR08 0.999327 0.998831 0.998471 0.998351* 0.00701 97783 RHACYR09 1.003938 1.002784 1.003235 1.002804 0.10683 151823 RHACYR08 1.001311 1.0 00327 1.001251 1.001381 0.12348 142784 MCRHAYR09 1.000134 1.000147 1.000173 1.000059 0.00130 5552 MCRHAYR08 1.000245 1.000255 1.000213 1.000227 0.00675 21255* PRICE 1.000018 1.00003** 0.000833* 1674* ADMS 1.00002*** 1.00002*** 1.00002*** 1.00002 *** 0.000059 1046*** ALOS 0.997224 0.94879** 0.95124** 0.94650*** 1.17205 809109 FP 1.01185 0.8802*** 0.86369*** 0.84318*** 2.83939*** 3936593** TEACH 0.98897 1.134849* 1.18934*** 1.160452** 4.27269* 26706752*** CASEMIX 2.79418*** 2.249572*** 2.361 176*** 2.7547*** 1.20803 3424433 TRANS 1.027049 1.044669 1.048667* 1.0521* 0.53814 1503427 HHI 0.886078 1.044669 1.064441 1.100902 4.84147* 3029843 MCAID 1.00971*** 1.004229* 1.004149 1.003747 0.02732 41695 INCOME 1.000003 1.000002 1.000002 1.00000 3 0.0002** 370 ** OCCRATE 1.01324*** 1.0114*** 1.00912*** 1.01065*** 0.15962*** 33815 BEDSIZE 1.32956*** 1.3295*** 1.32267*** 1.34059*** 0.10121 3628963*** POPDEN 0.999994 0.999983 0.999989 0.999983 0.00002 149 MDPOP 0.996068 0.972953* 0.979454 0.9 75807* 0.36223 2047253** PPERS 1.000004*** 1.000004*** 0.000021 36.99 PSUPP 1.000 1.000 0.000045 8.23 INTERCEPT 46233727*** 76359235*** 77009388*** 41951607*** 0.91772 12336315*** Key: *** means 1% significance, ** means 5% significance and means at 10% significance

PAGE 124

124 Table 5 11 Results for main effects model of HAC (hospital acquired conditions present or not present on admission) on financial performance of acute care hospitals Variable LDIREV LREV LCOST LDICOST MARG PROFIT HAC 0.99 0.98 0.97*** 0.97*** 0.24568 226900 MCARE 0.99 0.99*** 0.99*** 0.99*** 0.00204 185127* YR09 1.00 1.03 0.99 1.00 2.69598*** 5413841*** YR08 0.97 0.99 0.98 0.98 0.40610 1017070 PRICE 1.00* 1.00*** 0.00094* 1800** ADMS 1.00*** 1.00*** 1.00*** 1.00*** 0.00005 1037*** ALOS 1.00 0.96* 0.97 0.97 1.43570* 585542 FP 1.04 0.90*** 0.89*** 0.87*** 2.86717*** 4162004** TEACH 1.01 1.15** 1.21*** 1.18*** 4.32182* 27265500 *** CASEMIX 2.90*** 2.32**** 2.44* ** 2.87*** 1.87982 3392396 TRANS 1.03 1.05* 1.05** 1.06** 0.48333 1811874 HHI 0.91 1.06 1.07 1.10 4.92653* 3227754 MCAID 1.01*** 1.00 1.00* 1.00 0.02609 76831 UNEMP 0.99 0.99 0.98* 0.98 0.67006 408495 OCCRATE 1.01*** 1.01*** 1.01** 1.01*** 0.16247*** 53767 BEDSIZE 1.33*** 1.33*** 1.32*** 1.34*** 0.15980 3507342 *** POPDEN 1.00 0.99 0.99 0.99 0.0000272 4.54828 MDPOP 0.99 0.97* 0.98 0.98 0.37163 1958993* PPERS 1.00*** 1.00*** 0.0000122 23 PSUPP 1. 00 1.00 0.00004850 16 INTERCEPT 45240900*** 75078068*** 75859731*** 41399302*** 1.03119 11904210*** Key: *** means 1% significance, ** means 5% significance and means at 10% significance

PAGE 125

125 Table 5 12 Results for m ain effects of RHAC ( cate gorized into 4 groups) and MCARE (categorized into 4 groups) on financial performance of acute care hospitals Variable LDIREV LREV LCOST LDICOST MARG PROFIT Infection1 0.86*** 0.87*** 0.87*** 0.87*** 1.79641 2348012 Infection2 0.99 1.01 1.03 1.01 1.69 96 1524463 MCARE1 1.16** 1.31*** 1.29*** 1.28*** 0.32407 11305404*** MCARE2 1.05 1.13** 1.12*** 1.12*** 0.51192 5010709 MCARE3 1.05 1.09** 1.07** 1.08** 0.51833 972677 YR09 0.99 1.03 0.99 1.00 2.7763*** 5998339*** YR08 0.96 0.98 0.97 0.96 0.24609 9 02379 PRICE 1.00 1.00** 0.0009525* 2355*** ADMS 1.00*** 1.00*** 1.00*** 1.00*** 0.0000491 1015*** ALOS 0.99 0.95** 0.95*** 0.94*** 1.13548 1237855 FP 1.02 0.89*** 0.87*** 0.85*** 2.70503*** 4083133** TEACH 0.98 1.13* 1.19*** 1.16** 4.93238** 29 277460*** CASEMIX 2.82*** 2.25*** 2.39*** 2.80*** 1.15422 2533251 TRANS 1.02 1.04 1.04* 1.05* 0.72194 2356988 HHI 0.87* 1.02 1.05 1.09 4.835* 4832134 MCAID 1.01*** 1.00* 1.00** 1.00** 0.01028 242802 INCOME 1.00 1.00 1.00 1.00 0.0002** 366** UNEMP 0.99 0.98 0.98* 0.98 0.70502 57412 OCCRATE 1.01*** 1.01*** 1.01*** 1.01*** 0.14975*** 6629 BEDSIZE 1.33*** 1.32*** 1.31*** 1.33*** 0.1841 3419982*** POPDEN 1.00 1.00 1.00 1.00 0.00004 137 MDPOP 1.00 0.98 0.99 0.98 0.467 1434889 PPERS 1.00*** 1.00*** 0.00002 63 PSUPP 1.00 1.00 0.00005 39 INTERCEPT 45699104*** 70632041*** 71393215*** 39193550*** 0.51487 10229506** Key: *** means 1% significance, ** means 5% significance and means at 10% significance

PAGE 126

126 Table 5 13 Results for f ull mod el effects of categorized RHAC and MCARE on financial performance of acute care hospitals Variable LDIREV LREV LDICOST LCOST MARG PROFIT Infection1 0.96 0.95 0.95 0.98 2.91151 2154395 Infection2 1.02 1.05 1.03 1.05 1.64406 209133 Infection3 1.07 1.0 8 1.09 1.08 0.57462 1029120 MCARE1 1.21 1.34*** 1.29*** 1.28*** 2.64338 10473721* MCARE2 1.04 1.10 1.07 1.06 0 .94989 2355042 MCARE3 1.05 1.07 1.06 1.05 0.37766 3390927 YR09 1.13 1.11 1.06 1.03 4.24490 3317522 YR08 1.02 1.01 0.99 0.98 0 .39419 2966 15 INF1YR09 0.80* 0.84* 0.84* 0.85* 1.67488 89633 INF2YR09 0.93 0.96 0.98 0.98 0.20720 1139973 INF3YR09 0.85 0.88 0.88 0.90 1.42726 993137 INF1YR08 0.87 0.90 0.88 0.89 2.42535 594689 INF2YR08 0.94 0.93 0.93 0.93 0.28076 1886208 INF3YR08 0.90 0. 93 0.93 0.95 2.23830 1759441 MED1YR09 0.88 0.91 0.93 0.97 5.30079 8869 MED2YR09 1.01 1.04 1.07 1.07 0.34797 4868780 MED3YR09 0.99 1.04 1.03 1.04 0.83908 3652263 MED1YR08 1.01 1.02 1.03 1.03 0 .02536 2669892 MED2YR08 1.03 1.05 1.08 1.07 0 .10569 30 94808 MED3YR08 1.02 1.03 1.03 1.04 0 .83544 2753939 PRICE 1.00 1.00** 0 .00087351* 2376** ADMS 1.00*** 1.00*** 1.00*** 1.00*** 0 .00004777 1015*** FP 1.02 0.89*** 0.85*** 0.87*** 2.66500** 3995721** TEACH 0.97 1.12 1.16** 1.19*** 5.15050** 29189591 *** CASEMIX 2.89*** 2.27*** 2.82*** 2.39*** 1.27305 2219569 TRANS 1.03 1.04 1.05* 1.04* 0.93431 2515424 HHI 0.87* 1.02 1.09 1.05 5.17751* 5111289 MCAID 1.01*** 1.00** 1.01** 1.01*** 0.02709 258453* INCOME 1.00 1.00 1.00 1.00 0.0001898** 368** UNE MP 0.99 0.98 0.98 0.98* 0.79394 47210 OCCRATE 1.01*** 1.01*** 1.01*** 1.01*** 0.14715*** 2053 BEDSIZE 1.32*** 1.32*** 1.33*** 1.31*** 0.11556 3494346*** POPDEN 1.00 1.00 1.00 1.00 0.00000603 182 MDPOP 1.00 0.98 0.98 0.99 0.57447 1386069 PPERS 1.00*** 1.00*** 0.00002448 66 PSUPP 1.00 1.00 0.00004940 42 INTERCEPT 43049691*** 68138550*** 38034879*** 70121800*** 0.05180 11627863**

PAGE 127

127 CHAPTER 6 DISCUSSION This chapter begins with a brief summary of the dissertation, discusses the r esults for the patient outcome question and then concludes with a discussion of the financial performance question. Summary As part of the value purchasing program, Medicare implemented a policy that targeted 8 hospital acquired conditions that are reasona bly preventable and which normally resulted in higher reimbursement because of the higher DRG codes associated with the complications of treating them. This study evaluated the impact of this policy one year after its implementation. The two broad goals we re to evaluate the impact on patient outcomes and the impact on the financial performance of acute care hospitals. The study was designed to compare the probability of a patient acquiring these conditions before the policy was implemented with the probabil ity after the implementation and to compare hospital financial performance before and after the policy was implemented. Hypothesis: Patient Outcomes Hypothesis 1 : Medicare non payment policy will reduce the probability that an admitted patient acquires any of the hospital acquired conditions This hypothesis was supported by the results. Both the ANOVA and multivariate analysis showed a decrease in hospital acquired conditions (HAC) and hospital acquired infections (HAI). The reduction was greatest for HAI and further analysis shows that it was mainly due to a reduction in vascular catheter associated infections since cases of mediastinitis that meets the criteria for non payment were not found in the sample. The reduced probability in CAUTI was not statisti cally significant.

PAGE 128

128 Hypothesis 2 The reduction in probability will be higher for hospitals with higher previous year proportion of Medicare admissions This hypothesis was not supported by the results. The rationale for the hypothesis was based on the pres upposition that hospitals with a higher proportion of Medicare admissions in the previous year would have a greater incentive to improve on their infection control practices, in order to curtail the potential higher revenue losses. Patients in hospitals wi th more than 50% Medicare admissions were more likely to have these conditions than patients at hospitals with less than 50% Medicare admissions. However in both the ANOVA and the multivariate analysis, patients in hospitals with lower than 50% Medicare a dmissions had a higher reduction in the probability of acquiring these conditions post policy than those in hospitals with higher than 50% of Medicare admissions (although the findings were not statistically significant for CAUTI and HAI). Hypothesis 3 The reduction in probability will be higher for hospitals with lower previous year operating margins per admission This hypothesis was partially supported. The rationale was that hospitals with lower operating margin in the previous year would have a higher incentive to improve their infection control practices in order to mitigate the potential loss in revenue post policy. Patients in hospitals with higher operating margins were less likely to have CAUTI and HAC, but the finding was not statistically signifi cant for HAI. However, post policy, hospitals with lower operating margins were associated with statistically significantly lower probability of HAC than hospitals with higher operating margins but the findings were not statistically significant for HAI an d CAUTI. Hypotheses: Financial Performance Hypothesis 1 The higher the rate of hospital acquired conditions the higher the operating cost of hospitals. This hypothesis was supported by the results, which means that the

PAGE 129

129 higher the proportion of conditions not paid for by the policy the higher the expense of hospitals. This is in agreement with findings that the cost of treatment of these conditions is high. Hypothesis 2 Since Medicare used to pay higher DRGs for these conditions and other third parties have not announced cessation of these payments, the higher the rate of these hospital acquired conditions, the higher the operating revenues of hospitals. The results support this hypothesis in that a 1% increase in hospital acquired conditions was associated with about a 0.5% increase in revenue for hospitals. This confirms the assertion by Medicare that these conditions result in DRGs with higher reimbursement rates. Hypothesis 3 There will be a reduction in inpatient care revenue post policy the higher the n umber of hospital acquired conditions not paid for by Medicare This hypothesis was not supported by the results, which may be due to reductions from hospital acquired conditions being too small to influence the overall revenue received. Hypothesis 4 The reduction in inpatient care revenues will be higher for hospitals with higher proportion of Medicare admissions This hypothesis was not supported by the results. There were no statistically significant differences between hospitals with higher Medicare ad missions post policy and pre policy. In the ANOVA analysis, hospitals with lower Medicare admissions (below 50%) had higher revenues than those with Medicare admissions above 50% in 2009 compared with revenues in 2008. However, in the multivariate analysis these differences are explained away by the control variables. Hypothesis 5 There will be an increase in inpatient care expenses post policy the higher the number of hospital acquired conditions not paid for by Medicare. As expected, the higher the propo rtion of the hospital acquired conditions the higher the patient care expenses.

PAGE 130

130 However this hypothesis was not supported by the results because these higher inpatient care expenses were not statistically significant in any of the models. Hypothesis 6 The increase in inpatient care expenses will be higher for hospitals with higher proportion of Medicare admissions This hypothesis was partially supported by the results, at the 10% level of significance. However, in a subsequent analysis that categorized inf ection and Medicare, the proportion of Medicare admissions ceased to be statistically significant. Hypothesis 7 Profit margins will be lower for hospitals with higher proportion of Medicare admissions post policy and lower for hospitals with higher rate of hospital acquired conditions not paid for by Medicare post policy This hypothesis was not supported by the empirical results. Although hospitals generally had higher operating margins in 2009 than they did in 2007 and 2008, there were no significant diff erences post policy depending on the level of Medicare admissions of a given hospital. Other Interesting Findings The study identified certain patient characteristics that increase the risk of acquiring one of these conditions. These conditions included se x (females more likely than males), age (older you are the more likely you are), race (whites were more likely than the other races) and type of insurance (Medicare holders were less likely to acquire these conditions). Commercial insurance holders were fo und to be between 1.6 to 2.5 times the odd of Medicare patients in acquiring these hospital acquired conditions. Medicare policy indirectly affects all patients and there is no evidence that post policy hospitals were paying more attention to Medicare pati ents than other patients. It could also be that because hospitals are not reimbursed by Medicare for these conditions, but at the time of the study most commercial insurers did reimburse them, hospitals did not have the incentive to be vigilant on coding M edicare cases as compared to commercial

PAGE 131

131 insurance cases. However some commercial insurers like Aetna, Cigna, and Blue Cross Blue Shield are reported to have started implementing simi lar reimbursement reforms (NCSL 2010).

PAGE 132

132 CHAPTER 7 CONCLUSION The impac t of Medicare policy has been of interest to researchers over the years because Medicare is a major payer and payment policies affect the structure and conduct of the health care market. Previous policies have had lasting impacts, as discussed earlier, but skepticism surrounded the possible impact of this non payment policy on hospital behavior, due to the paucity of these conditions in discharges and the possibility of mitigating the impact through the present on admission indicator. However, I did not fin d any evidence of present on admission indicator abuse. Some researchers (McNutt et al 2009, McNair, Luft, and Bindman 2010, and Meddings Saint, and McMahon 2010) used simulations to predict the possible impact of this policy on and generally predicted that, because of the possible small financial impact, the policy might not affect hospital behavior. The goal of t his study was to use data pre and post policy to evaluate the impact of this policy first on patient outcomes and the n on the financial performance of acute care hospitals. The empirical findings indicate that the policy has not had a significant impact on catheter associated urinary tract infections (with the exception of hospitals with less than 25% Medicare admissions ), but were associated with a slight reduction in the probability of acquiring hospital acquired conditions collectively. The biggest impact of the policy so far is the significant reduction in the probability of a patient acquiring at least one of the thr ee hospital acquired conditions. Further analysis singled out vascular catheter associated as the most improved of the three conditions. The federal register for 2008 reported the average charges per hospital stay submitted by hospitals in 2007. Hospitals submitted $44,043 for CAUTI, $103,027 for VCAI, and $299,237 for Mediastinitis. From these figures, it is reasonable to assume that it costs

PAGE 133

133 hospitals more than twice as much to treat VCAI than CAUTI, and that could account for the significant improvement seen in VCAI. As a reminder, there were no cases of mediastinitis in the sample (MEDPAR reported only 69 cases natio nwide in 2007 (Federal R egister 2008)) Financially, the policy appears to have had little impact on acute care hospitals. The only significa nt effect found in this study was the increased inpatient care expenses post policy, which could mean that hospitals are responding positively to the policy by investing more on patient care. However, further research is needed to understand the source of the increased expenditure post policy. Limitations of the Study The validity of the study is predicated on the near perfect coding of both ICD 9 CM of these conditions and the correct use of the present on admission indicator. This study used discharge cla ims data but Zhan et al (2009) found discrepancies between medical claims coding and its corresponding condition from data extracted from medical records in New York and California. Secondly, data was from Florida which has a higher proportion of for prof it hospitals and therefore the findings cannot be extrapolated nationally. Thirdly, this is not a randomized control study and therefore it is subjected to all the limitations of an observational study. Causality is not implied and all findings are associa tions predicted ceteris paribus Fourthly, because the Medicare policy affected almost all hospitals in my sample, I could not use a control group. Further research could use a control group, for example federal hospitals like the Veterans Administration hospitals or critical access hospitals (these were exempt from the policy).

PAGE 134

134 Finally, these results should be seen as preliminary since we have just a year of data after the policy was implemented. In the keystone study, infection rates reduced to near zero within 6 months and the results were sustained even 18 months after the study. However, hospitals in that study had a higher motivation to reduce the infection rates because of the training and intense monitoring demanded by the study protocol. As more co nditions are added it will be interesting to further learn the sustainability of the improvements seen one year into the policy and what the continual impacts will be going forward. Policy Implication s The improved probability of acquiring these infections and conditions show that the policy is having a positive impact on hospital behavior. The financial impact has been little and although there is evidence that patien t care cost has increased post policy, it did not affect operating margins. There has been little improvement with CAUTI but significant improvement with VCAI and hopefully the trend might continue since Medicare keeps adding more conditions to the list. The financial incentive might be too little to elicit a drastic change like what happened i n the keystone project in Michigan however the increased awareness and maybe altruism could have been the driving force for this initial improvement. depending on the size of the incentive (Y oung 2005). As mentioned earlier it cost more than twice as much to treat VCAI than CAUTI and hospitals have drastically improved VCAI. CAUTI is the most prevalent of all the conditions but because it does not cost that much in treating it, hospitals might have strategically chosen the more costly conditions. Negative incentives are associated with behavior modification depending on the size, therefore as Medicare continues to expand these non payment conditions, there is the need to targ et conditions that are high cost and likely to attract the attention of hospital administrators.

PAGE 135

135 Future Research Further research is needed to understand what caused the increased in patient care cost, this will increase our understanding of the impact of the policy on hospital behavior. Secondly there is the need to further understand the managerial and cultural practices of high achieving hospitals to aid in our efforts to improve the quality of care in our hospitals. Thirdly, alternative methodology shou ld be exploited including the addition of a control group. Finally, analysis using the hospital as the unit of observation should be explored to understand the characteristics of successful hospitals. Final Thoughts Medicare added more conditions in 2008 including; surgical site infections following bariatric and some orthopedic procedures, deep vein thrombosis or pulmonary embolism after knee surgery and some poor d iabetic control practices (NCSL 2010). As these list increases and as other third party pay ers cease paying for these conditions, the revenue loss will rise to the point where it could spur drastic changes in practices in all hospitals. It is incumbent on hospitals to keep making the effort to gradually change the culture of tolerating any form of practices that hampers progress in spite of the current little financial impact of the policy.

PAGE 136

136 LIST OF REFERENCES Agency for He alth Care Administration (AHCA). 2010 [ Assessed May 8, 2010 ]. Available at: www.floridahealthfinder.gov/Researchers/o rderData/order note.shtml. Aikens, L.H., S.P. Clarke, R.B. Cheung, D.M. Sloane and J.H. Silber of Hospital Nurses and Surgical Patient M ortality Journal of American Medical Association 290 (12) 1617 1623. Alexander, J.W. 1985. P rogress Ann als of Surg ery 201: 423 428. Anderson, D.J., K.B. Kirkland, K.S. Kaye, P.A. Thacker II, Z.A. Kanafani G. Auten, and D.J. Sexton. 2007. Under Resourced Hospital Infection Control and Prevention Programs: Penny Wise, Pound F oolish? Infection Control and Hospital E pidemiology vol. 28, no.7 Apisarnt hanarak, A., K. Thongphubeth, S. Sirinvaravong, D. Kitkangvan, C. Yuekyen, B. f fectivenes s of Multifaceted Hospital Wide Quality Improvement Programs Featuring an Intervention to Remove Unnecessary Urinary Catheters at a Tertiary Care C enter in Thailand Infection Control Hosp ital Epidemioogy 28(7) 791 8 Arrow, K .J. 1963 nty and the W elfare Economics of Medical C are The American Economic Review 53(5):941 973. Avery G and J. Schultz. 2007 Regulation, Financial Incentives, and the Production of Q uality Am erican J ournal of Med ical Qual ity 22(4):265 73 Bazzolli, G.J., H ospitals belonging t o Health Networks and S ystems Inquiry Fall 37:234 252 Bazzoli G J ., R.C. Lindrooth, R. Hasnain Wynia, J. Needleman. 2004 2005. The Balanced Budget Act of 1 997 and U.S. Hospital Operations Inquiry winter 41 (4):401 17. ontrol Qua lity and Safety in Health care 13:233 234 Burke J P ., 2004 ontrol A Problem for Pati ent S afety N ew England J ournal of Med icine 348:651 6 Hospital Acquired Conditions: The Whys and Wherefores Plastic Surgical Nursing 28: 3 158 161. Chulis G S ., 1991. Assessing H ospitals Med ical Care Rev iew 48(2):167 206

PAGE 137

137 Clement J P ., M.J. McCue, R.D. Luke, J.D. Bramble, L.F. Rossiter, Y.A. Ozcan, and C.W. Pai Strategic Hospital Alliances: Impact on Financial P erformance. Health Affairs Nov/Dec. 16 (6):193 203. Cleme nt J P R.C. Lindrooth A.S. Chukmaitov and H.F. Chen. 2007 Matter in Hospital Patient Safety?: A Study of Urban H ospitals. Medical Care Feb ; 45 (2):131 8 Collins, T., J. Herness, J. Martenas, and A. Roberson 2007. Medicar e Prospective Payment Before and After Implementation: A Review of Visits and Physical Performance among Medicare Home Health Patients after Total Knee Replacements Home Health Nurse Jun 25(6):401 7 Coulam R. and G. Gaumer.1992 Payment System: A Critical HealthCare Financing Review 1991 annual supplement, 45 72 Cutler D. 1995. utc Econometrica (63) No.1,29 50 Davis C and D.J. Rhodes. 1988 Impact of DRG on the Cost and Quality of Health C are in the United States. Health Policy 9(2):117 31 DeBrock, L. and R.J. Arnould.1992 Utilization C ontrol in HMOs. Quarterly Review of Economics and Finance 32(2),31 53 ting the Quality of Medical Care The Milbank Quarterly Vol 83 (4): 691 729. Dulworth, S., and B. Pyenson .2004. Healthcare Associated Infections and Length of Hospital Stay in Medicare Population. American Journal of Medical Quality : 19;121 Feder J J. Hadley and S. Zuckerman System affect H ospitals? New Engl and J ournal of Med icine 317(14):867 873 Federal Register. 2008 National Archive and Records Wednesday April 30. [ Assessed May 13, 2010 ] Avail able at: http://edocket.access.gpo.gov/2008/pdf/08 1135.pdf Fisher, C 1992 inancial P erformance under PPS, 1985 90 Health Care Financing Review vol. 14 Frick, A P. S.G. Martin and M. Shwartz 1985. M ix and Cost Differences between Teach ing and Non Teaching Hospitals Medical Care vol.23; 283 295 Fried man, B., and S. Shortell. 1988. Owned and Not Fo r Profit H ospi tals before and after Medicare Prospective P ayment Health Services Research 23 (June):237 67

PAGE 138

138 Gapenski, L, W.V. Bruce and B. Langled Orban.1993. P rofitability Hospital and Health Services Administration 38 :1 Glickman S.W. F. S. Ou, E. R. Delong, M. T. Roe, B. L. Lytle, J. Mulgund, J. S. Rumsfeld, W. B. Gibler, E. M. Ohman, K.A. Schulman, and E.D. Peterson. .2007. Pay f or Performance, Quality of Care, and O utcomes in Acute M yoca rdial I nfarction J ournal of A merican M edical A ssociation 297:2373 2380 Gold, M. Containment E fforts: 1971 1993 Health Care Financing Review Spring ; 14 (3) 183 225 Gr f o r P erformance I nitiatives on the Quality of Care D elivery Medical Care Research & Review 63(suppl.) 29s 48s Hadley, J., S. Zuckerman and L. Iezzoni. 1996 Financial Pressure and Competition: Changes in Hospital Effici ency and Cost Shifting Behavior Medical Care 34 ( 3 ): 205 219. Haley, R., D. Quade, H. E. Freeman, and J. V. Bennett Nosocomial Infection Control (SCENIC PROJECT): Summary of Study Design American Journal of Epidemiology 111(5): 472 485. Hernandez S R and A.D. Kaluzny Proposed R esearch Health Service Research Fall 18:3 Hillman A. L., K. Ripley, N. Goldfarb, I. Nuamah, J. Weiner, and E Lusk. 1998 Financial Incentives and Feedback Failure to Increase Cancer S creenin g in Medicaid Managed C are Am erican J ournal of Public Health 88:1699 1701. Hoerger, T. J 1991 Profit and Not for Profit Hospitals Journal of Health Economics 10:259 289. sement Rules for Hospital Acquired Infections: How Electronic Reports can Protect the Bottom Line [Accessed February, 20 2010]. Available at: http://www.life sciencewriter.com/Getting_Ahead_of_Hospital_Acquired_Infection.pdf Johnson J R. M. A. Kuskowski, and T. J. Wilt 2006. Systematic Review: Antimicrobial Urinary Catheters to Prevent Catheter Associated Urinary Tract Infection in H ospita lized P atients Ann als of Intern al Med icine ;144(2):116 26 Kahn K L ., E. B. Keeler, M. J. Sherwood, W. H. Rogers, D. Draper, S. S. Bentow, E. J. Reinisch, L. V. Rubenstein, J. Kosecoff, and R. H. Brook. Care before and after Implementati on of the DRG Based Prospective Payment S ystem J ournal of t he American Medical Association Oct. 17 ; 264 (15):1984 8 Kah n C N ., T. Ault, H. Isenstein L. Potetzm and S. Van Gelder, 2006. Snapshot of Hospital Quality R eport ing and P ay For P erformance under Medicare Health Aff air ( Millwood) Jan Feb ; 25 (1): 148 62

PAGE 139

139 Kane R.L T. Shamliyan C. Mueller S. Duval, and T. Wilt 2007 Nursing Staffing and Quality of Patient Care. Evidence Report/Technology Assessment No. 151 (Prepared by the Minnesota Ev idencebased Practice Center under Contract No. 290 02 0009.) AHRQ Publication No. 07 E005.Rockville, MD: Agency for Healthcare Research and Quality. Klevens R M, J. R. Edwards, C. L. Richards T. Horan R. Gaynes D. Pollock and D. Cardo .2007 Estimati ng Healthcare Associated Infections in U.S. Hospitals, 2002 Public Health Rep orts 122:160 166. Konetzka R T D. Yi E. C. Norton and K. E. Kilpatrick .2004 Changes on Nursing Home Staffing and D eficiencies Health Serv ice Res earch 39(3):463 88. Kulesher R R perationa l History and Impact on Health C are Health Care Manag ement (Frederick) Jan Mar ; 25 (1):53 63. Kunkel S., U. Rosenqvist, and R. Westerling. 2007 Imp ortant to the Process and Outcomes, an Empirical Study of 386 Hospital D epartments in Sweden BMC Health Service Research 7:104 Li P J. E. Scneider and M. M. Ward 2007. ospit al C onversion on Patient S afety Health Servic e Research Dec ; 42 (6 pt 1): 2089 108 Lindenauer, P K., D. Remus, S. Roman, M. B. Rothberg, E. M. Benjamin, A. Ma, and D. W. Bratzler 2007. I mprovement N ew Engl and J ournal of Med icine 356(5 ):486 496. Lindrooth R C, G. J. Bazzoli, J. Needleman, and R. Hasnain Wynia 1993 Changes in H o spital Reimbursement on Staffing Decisions in Safety Net and Non S afety Net H ospitals N ew England J ournal of Med icine 329:1784 1789 Luft H S. J. P. Bunker and A. C. Enthoven .1979 perations be Regionalized? The Empirical R elati on between Surgical Volume and M ortality N ew Engl and J ournal of Med icine Dec. 20 ; 301 (25).1364 1369 McGiffin, D C D. C.Naftel, B. Czerska, R. S. D. Higgins, E. K. Kasper, O. H. Frazier, K. Aaronson, S. Davis, K. Thomas, R. C. Bourge 2002 The Impact of ERA and Co Morbidity on the Incidence of Infection after Cardiac Transplantation: a 10 Year Multi Institutional Study The Journal of Heart and Lung Transplantation 21: 56 57 McNair P., H S Luft, and A B Bindman Hospital Acquired Conditions: The Impact Health Affairs 28 ( 5 ). McNutt, R., T. J. Johnson, R. Odwazny, Z. Remmich, K. Skarupski, S. Meure r, S. Hohmann, and B. Harting. 2010. Change in MS DRG Assignment and Hospital Reimbursement as a Result of Centers for Medicare & Medicaid Changes in Payment for Hospital Acquired Conditions: Is it Coding or Quality? Quality Management in Health Care 19 (1)

PAGE 140

140 Meddings J., S. Saint, and L. McMahon Jr. 2010. Acquired Catheter Associated Urinary Tract Infection: Documentation and Coding Issues May Reduce Financial Impact Infection Control and Hospital Epidemiology June 31(6) Mehrotra A, C. L. Damberg M. E. Sorbero, S. S. Teleki. 2009. Pay for Performance in the Hospital Setting: What is the State of the E vidence? Am erican J ournal of Med ical Qual ity Jan Feb(1):19 28 Miller J T Y. R. Scott and M. Lee 2005 to the Development and S u ccess of Brain Tumor O perations Neurosurg ical Focus 18:E4 National Conference of State Legislatures Medicare Nonpayment for Hospital Acquired Conditions [Accessed on Nov., 28, 2010]. Available at: www.ncsl.org/default.aspx?tabid=14747 Newhouse, J.P. 1973. Towards a Theory of Nonprofit Institutions American Economic Review 60:87 92 Petersen, L. A., L. D Woodard, T. Urech C. Daw, and S. Sookanan. 2006 ay f or Performance Improve the Q uality of Healthcare? Annals of Internal Medicine 145:265 272 Phillips R L G. E. Fryer F. M. Chen S. E. Morgan L. A. Green E. Valente and T. J. Miyoshi. 2 004 The Balan ced Budget Act of 1997 and the F inanci al Health of Teaching H ospitals Ann als of Family Med icine 2(1):71 8. Prendergast, C .1999. The Provision of Incentives in Firms Journal of Economic Literature 37 : 7 63 Pronovost, P., D. Needham, S. Berenholtz, D. Sinopoli, H. Chu, S. Cosgrove, B. Sexton, R. Hyzy, R. Welsh, G. Roth, J. Bander, J. Kepros, and C. Goeschel .2006. An Intervention to D ecrease Catheter Related Bloodstream Infections in the ICU N ew E ngland J ournal of M edicine (26) 355:2 725 2732 Rand C ompare. for P erformance [ Acc essed Feb 20, 2010 ] Available at: www.randcompare.org/analysis of options/ anlysis of hospital pay for performance Raudenbush, S. and A. Bryk 2002. Hierarchical Linear Models: Applications and Data Analysis Methods London: Sage Publications. Riley, G ., and J Lubitz l Volume and M ortality Health Care Financing Review 7(1):37 47

PAGE 141

141 Salzer M S., C. Nixon, L. Schut, A. James, M. S. Karver, and L. Bickman 1996. Relationship between S tru cture, Process, and Outcome: A Conceptual Framework for Evaluating Quali ty [ Acc essed on Feb.,20, 2010] Available at : http://www.eric.ed.gov/ERICDocs/data/ericdocs2sql/content_storage_01/0000019b/80/19/ ab/52.pdf Sch neider, H., and A. Mathios. 2006. Principal Agency Theory and Health Care Utilization Economic Inquiry (44) 429 441 Schneider J. E., R. L. Ohsfeldt, M. A. Morrisey, P. Li, T. R. Miller, and B. A. Zelner. 2007. inancial Performance of General H ospitals, 1997 2004 Inquiry 44:321 334 Scott, R D .2002 irect Medical Cost of Healthcare Associated Infections in U.S. H ospitals and the Benefits of Prevention.CDC Division of Healthcare Quality Promotion [Acc e ssed at 02/20/2010]. Available at : http://www.cdc.gov/ncidod/dhqp/pdf/scott_costpaper.pdf Sear, A M 2004. Effects of Medicare BBA spending Reductions on the P rofitability of Genera l Acute Care H ospitals Journal of Healthcare Management 49:2 March/April Seshamani, M., J. S. Schwartz, and K. G. Volpp. 2006 R eim bursement on Hospital M ortality Health Service Research 41:683 700 Shen he Effect of Financial Pressure on the Quality of Care in H ospitals Journal of Health Economics 22: 243 269 Shortell S M and J. P. LoGerfo. 1981. Hospital Medical Staff Org anization and Quality of Care: Results for Myocardial Infarction and A ppende ctomy Medical Care 19(10): 1041 1055 Simkin, J 1997 Spartacus Educational. Biography of Florence Nightingale [A cc essed on May 12, 2010 ]. Available at: http:// www.Spartacus.schoo lnet.co.uk/REnightiingale.htm Steinwald, B and L. A. Dummit 1983 Hospital C ase M ix C hange: S icker P atients or DRG C reep? Health Affairs 8(2):35 47. Tennyson D H 2000 Does System Membership Enhance Financial Performance in Hospitals? Medical Care Research and Review 57; 29 Thompson J M, and M. J. McCue .2010 H o spitals before and after the R ehabilita tion Prospective Payment S ystem Health Care Manage ment Re search Jan Mar;35(1):36 45. U.S Bureau of Labor and S tatistics. [Accessed on September 29, 2010] Available at : http://data.bls.gov/map/servlet/map.servlet.MapToolServlet?state=12&datatype=unemploy ment&year=2009&period=M12&survey=la&map=msa&seasonal=u

PAGE 142

142 U.S. Department of Health and Human Services. Center for Medicare and Medicaid Service s (CMS). 2006. Never Events [Assessed on September 29, 2010]. Available at : http://www.cms.hhs.gov/apps/media/pres s/release.asp?counter=1863 U. S. Department of Health and Human Services. Centers for Medicare and Medicaid Services Roadmap for Implementing Value Driven Healthcare in the Traditional Medicare Fee for Service Program [ Assessed on Feb. 1 2, 2010 ]. Available at : http://www.cms.hhs.gov/qualityinitiativesGenInfo/downloads/VBPRoadmap OEA_1 16_508.pdf U.S. Department of Health and Human Servi ces. Centers for Medicare and Medicaid (CMS). Hospital C ompare 2010. [A ccessed on Feb., 20, 2010]. Available at : www.hospitalcompare.hhs.gov/Hospital/search/welcome.asp ? Virksti s, K. L., J. Westheim, C. Boston Feischhauer, P. N. Matsui, and T. Jaggi. 2009 Safeguarding Quality: Building the Business Case to Prevent Nursing Sensiti ve Hospital Acquired Conditions The Journal of Nursing Administration 39 ( 7/8 ): 350 355 Wilenski G. 2000. The Balanced Budget Act of 1997: A Current Look at its Impact on Patients and P roviders. [ Acc essed on Jan. 29, 2010 ]. Available at : http://pages.stern.nyu.edu/~jasker(BBAl.pdf World Health Organization (WHO). 2002. Report on Nosocomial Infection [ Acc essed on May 5, 2010 ]. Available at ; http://www.who.int/csr/resources/publications/whocdscsreph200212.pd f Young G. J .2005. for Quality P rograms American Journal of Medical Quality 20 ; 144 Younis M Z. 2006 etween the Balance d Budget Act and Hospital P rofitability J ourna l Healthcare Finance spring : 32 (3):72 9 Younis, M. Z., H. Z. Younies, and F. Okojie 2006. erformance in t he United States of America: A Follow Up S tudy Eastern Mediterranean Health Journal Vol. 12(5). Zhan C A. Elixhauser, C. L. Richards Jr., Y. Wang, W. B. Baine, M. Pineau, N. Verzier, R. Kliman, and D. Hunt. 2009. Identification of Hospital Acquired Cat heter Associated Urinary Tract Infections f rom Medicare Claims Medical Care Vol. 47, #3 Zinn J ational Structure and the Delivery of Primary Care to O lder Americans Health Service Research 33 (2 pt. 2): 354 380.

PAGE 143

143 BIOGRAPHICAL SKETCH Samuel Kwame Peasah was born in Ghana, West Africa. A fter completing a B achelor of pharmacy degree in 1993 at the Un iversity of S cience and Technology in Ghana and practicing for about 7 years, he relocate d to the United States with his family. While working as a pharmacist at Walgreens he enrolled and graduated with an MBA in 2005 at University of Florida. Being inspired by the impact of policy on patient outcomes during the implementation of Medicare part D, he went back to University of Florida in 2007 for his PhD in health services research specializing in health economics. He successfully defended a diss ertation on the impact of Medicare nonpayment of nosocomial infection on patient outcomes and hospitals financial performance i n Florida and graduated in May 2011.