Recent Developments in the Analysis of Admissions Data

Recent Developments in the Analysis of Admissions Data Jon Bumbaugh and Christie Teigland The Thirteenth Population Health & Care Coordination Colloquium March 13 15, 2013

Discussion Objectives Introduction Focus on Hospital Readmission Rates: Background Hospital Readmission Measure Methodologies Medicare Advantage Data Available from the MORE² Registry Update on Medicare Advantage Hospital Readmission Rates Comparing Methodologies for Readmission Measures Currently in Use National Quality Forum Request for NCQA and Yale to Harmonize Measures The Debate Regarding Readmission Measure Definitions Continues. 2

Introduction About Inovalon Origins: 1998 Headquarters: Outside Washington, D.C. Employees: Approximately 1,500 office based personnel, and thousands of in community clinical and operations personnel Clients: Hundreds of Health Plans Integrated Healthcare Delivery Systems Regulatory Organizations Academic Institutions Physician Associations Pharmaceutical Companies Adoption: Solutions touch more than: 120 Million Members; 540,000 Physicians; and 220,000 Clinical Facilities in 99% of all U.S. Counties. Empowered: Leveraging the data insights of more than 6 billion medical events 3

MORE² Registry Medical Outcomes Research for Effectiveness & Economics Inovalon has compiled healthcare datasets which are among the most extensive and broadly representative in the marketplace. The MORE² Registry contains de identified health data for: More than 6.3 billion medical events, 541,000 physicians, 220,000 clinical facilities, and 86 million unique members All payer types: Commercial, Medicare, Medicaid 4

Add Osteo # s MORE² Registry Sample Counts: Patient Disease Groups through 2012 Alzheimer s Disease/Dementia: 1,019,996 Hepatitis B: 251,818 Anemia: 9,386,721 Hepatitis C: 538,785 Anxiety: 6,202,332 HIV: 342,527 Asthma: 8,397,075 HPV: 670,625 Atrial Fibrillation: 2,048,651 Lung Cancer: 351,634 All Cancers: 4,513,604 Macular Degeneration: 1,123,542 Bipolar Disorder: 1,199,786 Multiple Sclerosis: 197,580 Breast Cancer: 851,900 Myocardial Infarction: 1,579,195 Cardiovascular disease: 5,744,520 Prostate Cancer: 769,664 Colon / Colorectal Cancer: 445,971 Psoriasis: 769,664 COPD: 4,448,365 Pulmonary Embolism: 382,726 Diabetes Type I: 1,682,736 Rheumatoid Arthritis: 947,547 Diabetes Type II: 7,882,106 Schizophrenia: 464,579 Depression: 6,481,150 Stroke Hemorrhagic: 199,832 Dyslipidemia: 17,581,747 Stroke Ischemic: 966,029 5

Focus on Hospital Readmission Rates Background: Hospital readmissions are receiving increasing attention as a largely avoidable source of poor quality care and excessive spending. In April 2009, a study on readmissions within Medicare s fee for service (FFS) program found a 19.6% 30 day readmission rate in 2004. 1 Although studies have shown that specific interventions particularly for patients with multiple conditions reduce readmission rates by up to 50%, CMS found that Medicare s national 30 day readmission rate did not change appreciably between 2004 and 2009. 2 In 2010, the readmission rate for Medicare beneficiaries remained flat at 19.2%. 3 Just under 10 million admissions Approximately 1.9 million readmissions Cost estimated to exceed $17 billion (inpatient spending) 1 Jencks SF, Williams MV, and Coleman EA. Rehospitalizations among patients in the Medicare fee for service program. NEJM 2009; 360: 1418 28. 2 Berenson R, Paulus R, Kalman N, Medicare s Readmissions Reduction Program A Positive Alternative, NEJM, 2012; 366;15. 3 National Medicare Readmission Findings: Recent Data and Trends, Office of Information Products and Data Analytics, CMS presentation. 6

Focus on Hospital Readmission Rates Background: The Patient Protection and Affordable Care Act requires Medicare to penalize hospitals starting October 2012 by adjusting DRG payments for high readmission rates (i.e., worse than expected ) related to myocardial infarction, pneumonia, or heart failure. 4 Penalties of up to 1% were charged to 2,211 hospitals in the first year. Penalties will rise to up to 2% in FY 2014 and up to 3% in FY 2015. Beginning FY 2015, CMS plans to expand the program to include readmissions for other common diagnoses. Hospital readmission rates are publicly reported as a measure of quality. Crediting this program, CMS reported the first observed decrease in the rate of rehospitalizations in the final quarter of 2012. 5 After fluctuating between 18.5% and 19.5% from 2007 to 2011, the 30 day all cause readmission rate dropped to 17.8% in 2012. 4 medicareadvocacy.org http://www.medicareadvocacy.org/2012/05/02/medicare hospital readmissions/ accessed on 9/18/2012 5 http://www.healthleadersmedia.com/print/hep-28972130day-readmission-rates-fell-in-2012 accessed on 3/1/2013 7

Medicare Advantage (MA) Hospital Readmission Rates Using Inovalon s MORE² Registry An Inovalon study based on 2.4 million individuals drawn from 11 representative MA health plans published in AJMC in 2012 found the 30 day readmission rate was about 14.5% annually during the 2006 2008 period approximately 26% lower than the 19.6% FFS rate reported by Jencks. 6 Patients at Risk at Beginning of Period Cumulative Readmissions 0 30 Days After Discharge Percent Readmitted FFS (Jencks et al) 2004 2,961,460 579,903 19.6% MA 2006 62,012 9,069 14.6% 2007 75,847 11,013 14.5% 2008 93,226 13,559 14.5% *In all years, rates were significantly (p<.001) less than the Medicare FFS rate. 6 Lemieux M, Sennett C, Wang R, Mulligan T, Bumbaugh J. Hospital Readmission Rates in Medicare Advantage Plans, Am J Managed Care. 2012;18(2):96 104 8

Hospital Readmission Measures The Impact of Definitions & Methodologies Different definitions of number of readmissions and number of discharges and other methodological differences result in a divergence of rates across various measures currently in use. Considerations: 1. How is the universe of patients defined? 2. What types of cases are excluded from index admissions? 3. What types of cases are excluded from countable re hospitalizations? e.g., does the definition distinguish and exclude planned surgical, medical or other planned admits? 4. What risk adjustment methodology is used? 5. How are multiple readmissions for same patient counted? 6. Are only clinically related admissions considered? 9

Hospital Readmission Measures Definitions & Methodologies Jencks definition: Defines a hospital readmission as an admission to a hospital within 30 days after discharge following an original (index) admission and discharge. Numerator = Denominator = number of re hospitalizations number of index discharges in given period of time Calculates rates over a 12 month period for the cohort discharged during last quarter of given year (October 1 through December 31 in prior year). Counts only the first readmission for each discharge. Excludes patients who were transferred on day of discharge to another acute care hospital. Excludes patients re hospitalized for rehabilitation within 30 days after discharge. Excludes data for patient if death occurred. (Note: at the health plan level, disenrollment from the plan is also a censoring event) 10

Update on Medicare Advantage Hospital Readmission Rates As of January 2013, approximately 14 million Medicare beneficiaries were enrolled in a Medicare Advantage (MA) health plan. 7 MA plans have substantial flexibility to arrange coverage and develop networks of health care providers to serve their enrollees. Most MA plans are local health maintenance organizations (HMOs), also called coordinated care plans (CCPs) or regional preferred provider networks (PPOs). Because MA plans are paid on a capitated basis (a fixed, risk adjusted amount per enrollee), they have added incentive to attempt to reduce costly avoidable hospitalizations and readmissions via case management or network contracting arrangements. 7 Centers for Medicare and Medicaid Services (CMS), Monthly Enrollment by Contract (January 2013). Includes COST, PACE and DEMO enrollees. Data download available at: https://www.cms.gov/research-statistics-data-and-systems/statistics- Trends-and-Reports/MCRAdvPartDEnrolData/Monthly-Enrollment-by-Contract-Items/Enrollment-by-Contract-2013-01.html 11

Update on Medicare Advantage Hospital Readmission Rates Using the MORE² Registry Where is the MA 30 day readmissions rate now? Inovalon updated the earlier study using the MORE² Registry to see if rates have improved since 2008 when CMS began targeting readmission rates. 2008 2009 2010 2011 Unique members 6,183,167 6,685,043 6,332,116 5,015,060 Average membership 4,802,403 4,946,185 4,636,511 3,409,665 Member months 57,628,839 59,354,216 55,638,138 40,915,978 Members with at least one discharge 802,123 812,694 679,457 491,378 Percentage of members with at least one discharge 14.6% 14.5% 12.9% 11.5% Number of discharges 1,323,496 1,284,362 1,052,595 733,948 Annual discharge rate per 1,000 members Population Characteristics 276 260 227 215 Number of hospital days 7,739,633 7,864,821 6,449,788 4,508,464 Average length of hospital stay 5.8 6.1 6.1 6.1 12

Medicare Advantage Plan Demographics MORE² Registry Age Distribution Age distributions are similar and stable across years. MORE 2 Registry (Medicare Advantage) CMS Denominator File 5% Sample (Medicare Advantage) 13

Medicare Advantage Plan Demographics MORE² Registry Gender Distribution Gender distributions are similar and stable across years. MORE 2 (Medicare Advantage) CMS Denominator File 5% Sample (Medicare Advantage) 14

Medicare Advantage Plan Demographics MORE² Registry Geographic Distribution Geographic distributions are similar and stable across years. MORE 2 Registry sample has slightly larger representation in Northeast and slightly lower representation in the South and West. MORE 2 Registry vs. CMS Denominator File 5% Sample Medicare Advantage Members Census Region 2008 2009 2010 2011 MORE 2 MA 5% MORE 2 MA 5% MORE 2 MA 5% MORE 2 MA 5% Northeast 21.8% 21.2% 26.1% 21.0% 24.7% 20.7% 26.2% 20.4% Midwest 21.9% 17.8% 23.4% 18.1% 20.8% 17.6% 22.5% 18.4% South 34.9% 28.6% 28.6% 29.5% 30.9% 30.2% 26.2% 29.6% West 19.5% 28.2% 19.9% 27.3% 21.5% 27.1% 22.1% 27.1% US Territory 1.9% 4.2% 2.0% 4.1% 2.1% 4.3% 3.0% 4.6% 15

Update on Hospital Readmission Rates Using the MORE² Registry Percent Year Starting in 2009, we see the first signs of the rate declining, dropping to 13.8% in 2011. Medicare Advantage plans demonstrated reductions in readmission rates years sooner than the first sign of decline in Medicare Fee for Service (FFS) plans noted in 2012. 16

Update on Hospital Readmission Rates Using the MORE² Registry The 31 60 day readmission rates decline even more from 2008 to 2011. 17

Update on Hospital Readmission Rates Using the MORE² Registry And an even larger decline in the 60 90 day readmission rates. 18

Medicare Advantage Readmission Rates: 2011 Table 1. 2011 Rehospitalizations after Discharge from the Hospital among Patients in Medicare Advantage Programs Interval after Discharge Patients at Risk at Beginning of Period % Cumulative Readmission s by End of Period % Cumulative Death w/o Readmission by End of Period % Cumulative Disenrollmen t w/o Readmission by End of Period % 0 30 days 186,839 100.0% 25,789 13.8% 4,103 2.2% 13,170 7.0% 31 60 days 143,777 77.0% 37,488 20.1% 5,247 2.8% 26,517 14.2% 61 90 days 117,587 62.9% 44,851 24.0% 5,928 3.2% 36,413 19.5% 19

Comparing Methodologies for Different Readmission Measures Currently in Use CMS Hospital Readmission Measures Reported On CMS Hospital Compare NCQA Plan All Cause Readmission Measure Used In CMS Five Star Rating System 23

The Jencks Readmission Framework: Understanding the Challenges Four categories of readmissions: 8 1.Related and Unplanned. Readmissions that are related to index admission but not planned. For example, a person readmitted to address an adverse event caused by an infection or sepsis, or a person with heart failure who is readmitted for chest pain. 2.Related and Planned. Readmissions that are related to the initial hospitalization and are scheduled in advance to deliver follow up care and/or perform medical procedures. For example, a patient may be admitted for heart failure and readmitted later for the placement of a cardiac stent. 3.Unrelated and Planned. Readmissions that are unrelated and planned. For example, an admission for chronic obstructive pulmonary disorder (COPD) followed by a readmission for a scheduled hip replacement surgery. 4.Unrelated and Unplanned. Readmissions that are unrelated to the initial hospitalization and also unplanned. For example, readmissions for burns or traumas caused by accidents. Another example might be an initial admission for a gastrointestinal disorder followed by a readmission for skin cancer. 8 Stephen F. Jencks, M.D., M.P.H., Rehospitalization: Understanding the Challenge, Presentation at the National Medicare Readmissions Summit, Washington, DC, June 1, 2009. 24

CMS Hospital Readmission Measures 9 Horwitz L, et al, Hospital-Wide All-Cause Unplanned Readmission Measure, Final Technical Report, Yale New Haven Health Services Corporation/Center for Outcomes Research & Evaluation, Submitted to CMS, July 2012. 25

CMS Hospital Readmission Measures All three condition specific hospital measures exclude certain events: Patients who die during the initial hospitalization. Same day readmissions to same hospital for same condition. Patients who are transferred out of the hospital to another acute care facility. Patients who are discharged against medical advice. Specific planned admissions (e.g., amputations, hip replacement, mastectomy, kidney transplant, other organ transplant). 9 Only the AMI measure excludes certain planned follow up procedures. The AMI measure also excludes patients who are discharged on same day of admission. CMS did not exclude any planned readmissions for heart failure or pneumonia because clinical experts who were consulted did not identify common follow up causes for a scheduled procedure that would represent continuing treatment care for these conditions. 9 See Horwitz L, et al, Hospital-Wide All-Cause Unplanned Readmission Measure, Final Technical Report, p. 13-14 for full list of conditions considered planned. 26

CMS Hospital Readmission Measures Debate on Related vs. Unrelated Readmissions CMS position is that determining whether the readmission is related to the original admission cannot be made solely on the basis of the admitting diagnosis for the readmission. 10 Arguments: Limiting readmissions to particular diagnoses would permit a hospital to avoid countable readmissions by changing coding practices. It could also create an incentive to avoid patients with conditions that are part of readmission measures. Re hospitalizations that are not related to the original admission should not affect some hospitals disproportionately similar patients should have same probability of readmission. Hospital advocates maintain the small set of existing excluded readmissions does not meet the statutory requirement that unrelated readmissions not be counted and that hospitals will be penalized for readmissions beyond their control. Recommend excluding patients with conditions such as trauma, burns, substance abuse and psychiatric disorders. Recommend adding a claims modifier so a hospital can identify planned readmissions. 10 Federal Register, August 18, 2011, vol. 76, no 160, p. 51669. 27

CMS Hospital Readmission Measures The measures use the hierarchical generalized linear model approach. This technique accounts for the clustering of patients within hospitals based on the assumption that an individual hospital will provide similar quality of care across patients within its patient population, which can be measured using hospital specific intercepts. 11 The hospital specific intercepts account for the non independence of patients within the same hospital. If there were no differences among hospitals, after adjusting for patient risk, the hospital intercepts should be identical across all hospitals. This approach accounts for variation across hospitals in how sick their patients are when admitted to the hospital (use of AHRQ chronic condition categories for risk adjustment) to reveal differences in hospital specific quality. 11 Tilson, S and Hoffman, G, Addressing Medicare Hospital Readmissions, Congressional Research Service, Report for Congress, 7 5700, R42546, May 25, 2012. 28

CMS Hospital Readmission Measures Measure rate calculation: The predicted number of readmissions in each hospital is estimated given the same patient mix and its hospital specific intercept. The expected number of readmissions in each hospital is estimated using its patient mix and the average of each of the estimated hospital specific intercepts. The excess readmission ratio = predicted number / expected number of readmissions. The ratio is measure of relative performance: Hospitals performing better than the average hospital that admitted similar patients (based on patient risk factors and comorbidities), the ratio will be less than one. Hospitals performing worse than the average hospital after risk adjustment will have a ratio greater than one. NOTE: This methodology is more difficult to explain to the public and other stakeholders who are more familiar with the approach that uses an observed over expected ratio determined in a logistic regression model (e.g., NCQA all cause 30 day readmission measure). 9 Horwitz L, et al, Hospital-Wide All-Cause Unplanned Readmission Measure, Final Technical Report, Yale New Haven Health Services Corporation/Center for Outcomes Research & Evaluation, Submitted to CMS, July 2012. 29

NCQA Plan All-Cause Readmissions NCQA s measure is used in the FIVE STAR ratings for Medicare Advantage plans and will be factored into the mandated quality bonus payments under the Affordable Care Act. 12 Uses administrative claims to determine the number and percentage of acute inpatient stays during the measurement year followed by an acute readmission for any diagnosis within 30 days. Health plan members 18 years and older. Members must have been continuously enrolled in the plan for 365 days prior to the discharge date through 30 days after the discharge date with no more than one gap of 45 days or less within the 365 days prior to discharge date. Exclusions include maternity related stays; admission to a long term care facility; death during admission or readmission; admissions with same day discharge date. Measure covers all types of discharges, not just particular conditions as in hospital measures. Index Hospital Stay (IHS): An acute inpatient stay with a discharge on or between January 1 and December 1 of the measurement year. Acute inpatient stays include general medical and surgical hospital stays where the patient is discharged to a community setting. Index Discharge Date: The IHS discharge date. The index discharge date must occur on or between January 1 and December 1 of the measurement year. Index Readmission Stay: An acute inpatient stay for any diagnosis with an admission date within 30 days of a previous index discharge date. 12 2012 Insights for Improvement, Reducing Readmissions: Measuring Health Plan Performance, an NCQA Insights for Improvement Publication. 30

NCQA Plan All-Cause Readmissions Measure calculation: Numerator: Count of 30 day readmissions with admission date between January 2 and December 31 of measurement year Denominator: Count of index hospital stays Observed rate of readmission = Numerator/Denominator Risk Adjustment: To allow fair comparison among plans, the measure is risk adjusted via indirect standardization, using predicted probabilities of readmission estimated through logistic regression. Risk adjustment is applied by assigning a weight to each index hospital stay, based on the presence of surgery, discharge condition, comorbidity, age and gender. The Clinical Conditions (CC) and HCC categories identify comorbidities and attach weights to each statistically significant comorbidity by product line (i.e., commercial, Medicare) and age group (18 64 and 65 and older). Weights were developed separately for each product line using a testing database that includes members from multiple health plans. Expected rate of readmission (adjusted probability of readmission) is rate the plan is expected to have based on case mix and average health plans across the U.S. 31

NCQA Plan All-Cause Readmissions Observed to Expected Ratio (O / E) is calculated for health plan across all age and gender groups. When the ratio is <1.0, the health plan performed better than expected (as predicted by the model) When the ratio is >1.0, the plan performed worse than expected. In 2012, NCQA reported O/E ratios for commercial populations 18 65 years of age and for Medicare populations 65 years and older. 32

National Quality Forum Request for NCQA and Yale to Harmonize Readmission Measures 33

National Quality Forum (NQF) Requests for Yale / NCQA to Harmonize Respective Admission Measures 34

National Quality Forum (NQF) Requests for Yale / NCQA to Harmonize Respective Admission Measures Examples of key harmonization items requested by the NQF committee: 4.Allow readmissions to count as index admissions. NCQA modeled its approach to exclude readmissions as index events based on prior Yale work on the condition specific readmission measures and the standard used in the literature. NCQA rationale for approach: The intent of counting readmissions as index events is to hold hospitals accountable for the total impact of mistakes that lead to readmission and failure to correct on subsequent readmissions. However, this has the effect of double counting (or more) the impact of comorbid conditions in the risk model, which may lead to erroneous conclusions about which factors predict readmission. Example: a patient admitted for diabetes is discharged on a new regimen and is not adequately educated on selfmanagement. The patient becomes hypoglycemic two days later because of confusion between short and long acting insulin and is readmitted and experiences further complications that lead to a chain of < 30 day readmissions. Suppose this patient has a diagnosis of COPD recorded in preceding 12 months. Dx will count in the regression model for each hospitalization and strengthen the association between that condition and the readmission. This additional weighting of comorbid conditions will make it harder to identify new conditions like the hypoglycemia as predictors of readmission and may misdirect hospitals and plans quality improvement activities. 35

National Quality Forum (NQF) Requests for Yale / NCQA to Harmonize Respective Admission Measures 5. Allow readmissions to count as index admissions (continued) Yale rationale for counting readmissions as index readmissions: Institutions should be held accountable for all readmissions. A readmission is a signal that discharge planning may have been inadequate if further readmissions are not counted in measure, there is no incentive to prevent them. Some conditions/diagnoses may be more likely to be readmissions e.g., infections if we do not count as index case we may be excluding some conditions more than others. Status: NCQA is developing and testing a modified specification to assess the impact of counting readmissions as index admissions on plan performance relative to current specification. If the change results in an improvement and meets scientific acceptability criteria, NCQA would initiate the process to implement the change. 36

The Controversy Regarding Readmission Measure Definitions Continues. March 6, 2013: Joynt K, Jha A, A Path Forward on Medicare Readmissions. New England Journal of Medicine, Perspective. Cite overwhelming evidence two groups of patients are at particularly high risk for readmissions: those who have the most severe illnesses (because of their underlying condition); those who are socioeconomically disadvantaged. The current measure does not account for these factors, leaving hospitals that disproportionately care for the sickest and poorest patients at greatest risk for penalties. CMS data shows that two thirds of eligible U.S. hospitals were found to have readmission rates higher than the CMS models predicted, and each of these hospitals will receive a penalty. This is much higher than anticipated on the basis of CMS's previous public reports, which identified less than 5% of hospitals as outliers. <http://www.nejm.org/doi/full/10.1056/nejmp1300122?query=featured_home&> 37

The Controversy Regarding Readmission Measure Definitions Continues. Evidence that safety net institutions and large teaching hospitals are far more likely to be penalized. The program has potential to exacerbate disparities in care and create disincentives to providing care for very ill patients with complex health needs. Recommendations: 1. Adjust readmission rates for socioeconomic status 2. Weighting the penalties based on timing of readmissions readmissions within first few days after discharge may reflect poor care coordination or inadequate discharge planning (weight heavy) readmissions 4 weeks later are far more likely to be due to the underlying severity of disease (weight far less) 3. Give hospitals credit for low mortality rates (based on the fact these hospitals tend to have higher rates of readmission and get penalized more than hospitals with high mortality rates). 38

Discussion Summary Introduction Focus on Hospital Readmission Rates: Background Hospital Readmission Measure Methodologies Medicare Advantage Data Available from the MORE² Registry Update on Medicare Advantage Hospital Readmission Rates Comparing Methodologies for Readmission Measures Currently in Use National Quality Forum Request for NCQA and Yale to Harmonize Measures The Debate Regarding Readmission Measure Definitions Continues. 39

Discussion