Hospital-Wide (All-Condition) 30‐DayRisk-Standardized Readmission MeasureDRAFT Measure Methodology ReportSubmitted By Yale New Haven Health Services Corporation/Center forOutcomes Research & Evaluation (YNHHSC/CORE):Leora Horwitz, MD, MHSChohreh Partovian, MD, PhDZhenqiu Lin, PhDJeph Herrin, PhDJacqueline Grady, MSMitchell Conover, BAJulia Montague, MPHChloe Dillaway, BAKathleen Bartczak, BAJoseph Ross, MDSusannah Bernheim, MD, MHSElizabeth Drye, MD, SMHarlan M. Krumholz, MD, SMContract number: HHSM-500-2008-0025I/HHSM-500-T0001, Modification No. 000005Prepared For:Centers for Medicare & Medicaid Services (CMS)Last updated August 10, 2011DRAFT1

TABLE OF CONTENTSLIST OF TABLES. 4LIST OF FIGURES . 5GLOSSARY OF TERMS . 61. INTRODUCTION . 71.1Overview of Measure . 71.2Hospital-wide Readmission as a Quality Indicator . 71.3Approach to Measure Development . 82. METHODS. 92.1Overview . 92.2Outcome Definition . 102.2.1Planned readmissions . 102.2.2Thirty-day timeframe . 142.2.3All-cause readmission . 172.3Definition of Eligible Population . 172.3.1Grouping patients into clinically coherent discharge condition categories by usingAHRQ Clinical Classification System (AHRQ-CCS) . / exclusion criteria . 18Administrative Model Development . 202.4.1Data sources . 202.4.2Multiple models . 222.4.3Models are defined by care team (service line) or clinical coherence . 222.5Risk Adjustment . 242.5.1Complications of hospitalization . 242.5.2Case mix adjustment: risk variables . 252.5.3Service mix adjustment . 262.6Statistical Approach to Model Development . 262.6.1Models for each cohort of conditions . 262.6.2Hospital performance reporting . 282.6.3Creating interval estimates . 282.6.4Algorithm . 283. RESULTS . 304. SUMMARY . 32DRAFT2

5. REFERENCES . 336. APPENDIX – TABLES . 36Appendix A – Top 10 primary discharge diagnoses for planned procedures . 36Appendix B – List of AHRQ procedure CCs defining the surgical/gynecological cohort. 45Appendix C – Condition categories assigned to the medical cohorts . 48Appendix D – Conditions that are treated as complications if occurring during indexadmission. 54Appendix E –Candidate risk variables . 56Appendix F – Final risk-adjustment variables . 57DRAFT3

LIST OF TABLESTable 1 – Planned procedure list . 12Table 2 – Discharge condition categories considered acute or complications of care. 13Table 3 – Admissions, readmissions and mortality for the seven cohorts (2008) . 24Table 4 – Frequency of hospitals in each cohort (Number of hospitals 4919) . 31Table 5 – Number of hospitals based on number of models to which they contribute . 31Table 6 – C-statistic for each model . 31DRAFT4

LIST OF FIGURESFigure 1 – Graphical representation of time to readmission . 15Figure 2 – Inclusion and exclusion criteria* . 21Figure 3 – Distribution of 30-day hospital-wide readmission rates . 30Figure 3a –Unadjusted rate . 30Figure 3b –Risk-standardized rate . 30DRAFT5

GLOSSARY OF TERMSTermDefinitionIndex admissionAny eligible admission to an acute care hospital assessed inthe measure for the outcome (readmitted or not within 30days). (See Section 2.3.2)ReadmissionAn admission to an acute care hospital within 30 days ofdischarge from an acute care hospital. (See Section 2.2). Areadmission may in turn serve as an index admission.Planned readmissionAn intentional readmission within 30 days of discharge from anacute care hospital that is a scheduled part of the patient’s planof care. Planned readmissions are not counted as outcomes inthis measure (see Section 2.2.1).Discharge diagnosisICD-9 level code of the principal reason for hospitalization.Discharge conditioncategoryA group of related discharge diagnosis ICD-9 codes, asgrouped by the Agency for Healthcare Research and Quality(AHRQ) Clinical Classification Software (CCS). (See Section2.3.1).CohortA group of admissions for patients with related conditioncategories or procedure categories; this measure includesseven cohorts, each with its own risk model (see Section2.4.3).Measure populationThe full set of admissions eligible for inclusion in the measure.Procedure categoryA group of related procedure codes, as grouped by the Agencyfor Healthcare Research and Quality Clinical ClassificationSoftware (CCS) (see Section 2.4.3).Risk variableA variable in the risk-adjustment model intended to account forpatient comorbid conditions or age. A risk variable mayinclude multiple conditions. Each condition is a group of ICD-9diagnosis codes, as defined by the Centers for Medicare andMedicaid Services Condition Category groups (CCs) (seeSection 2.5.2).Case mixThe variation among hospitals in illness severity and age oftheir patients.Service mixThe variation among hospitals in the types of conditions theycare for and procedures they provide.DRAFT6

1. INTRODUCTION1.1 Overview of MeasureReadmission following hospitalization is a costly and often preventable event. During 2003and 2004, almost one-fifth of Medicare beneficiaries – over 2.3 million patients – wererehospitalized within 30 days of discharge.1 Jencks estimated that readmissions within 30days of discharge cost Medicare more than 17 billion dollars annually.1 A 2006Commonwealth Fund report estimated that if national readmission rates were lowered tothe levels achieved by the top-performing regions, Medicare would save 1.9 billionannually. 2Currently, the Centers for Medicare and Medicaid Services (CMS) publicly reports riskstandardized readmission rates for heart failure, pneumonia and acute myocardialinfarction. 3-9 CMS has also developed hospital readmission measures for stroke and forhip and knee replacement, and is developing them for chronic obstructive pulmonarydisease and vascular procedures. While it is helpful to assess readmission rates forspecific groups of patients, these conditions account for only a small minority of totalreadmissions.10 By contrast, a hospital-wide, all-condition readmission measure couldprovide a broader assessment of the quality of care at hospitals. Therefore, CMS hascontracted with Yale New Haven Health Services Corporation/Center for OutcomesResearch and Evaluation (YNHHSC/CORE) to develop a claims-based, risk adjustedhospital-wide readmission (HWR) measure for public reporting that reflects the quality ofcare for hospitalized patients in the United States.In this technical report we provide detailed information on the development of the HWRmeasure. Briefly, we developed the measure as an all-condition measure designed tocapture unplanned readmissions within 30 days of discharge. The HWR measurecomplies with accepted standards for outcomes measure development, includingappropriate risk adjustment and transparency of specifications. The measure includes alladmissions except those for which a subsequent readmission would not be considered aquality signal. The measure does not count planned readmissions in the measureoutcome, since they do not represent a quality signal. The overall risk-standardizedreadmission rate is derived from a composite of seven statistical models built for groups ofadmissions that are clinically related. The seven risk adjustment models will be tested forreliability in a split sample dataset combining two calendar years (2007 and 2008), and thestability of the measure over time will be tested using data from 2009. Although wedeveloped the measure using Medicare data, the measure will also be tested in andadapted for all-payer datasets.1.2 Hospital-wide Readmission as a Quality IndicatorHospital readmission, for any reason, is disruptive to patients and caregivers, costly to thehealthcare system, and puts patients at additional risk of hospital-acquired infections andcomplications. Readmissions are also a major source of patient and family stress and mayDRAFT7

contribute substantially to loss of functional ability, particularly in older patients. Somereadmissions are unavoidable and result from inevitable progression of disease orworsening of chronic conditions. However, readmissions may also result from poor qualityof care or inadequate transitional care. Transitional care includes effective dischargeplanning, transfer of information at the time of discharge, patient assessment andeducation, and coordination of care and monitoring in the post-discharge period.Numerous studies have found an association between quality of inpatient or transitionalcare and early (typically 30-day) readmission rates for a wide range of conditions. 11-18Therefore, while readmission rates would never be expected to be zero, variation inreadmission rates for a broad spectrum of conditions is related to quality of care.Furthermore, randomized controlled trials have shown that improvement in the followingareas can directly reduce readmission rates: quality of care during the initial admission;improvement in communication with patients, their caregivers and their clinicians; patienteducation; predischarge assessment; and coordination of care after discharge.20-35Evidence that hospitals have been able to reduce readmission rates through these qualityof-care initiatives illustrates the degree to which hospital practices can affect readmissionrates. Successful randomized trials have reduced 30-day readmission rates by 20-40%.Widespread application of these clinical trial interventions to general practice has alsobeen encouraging. Since 2008, 14 Medicare Quality Improvement Organizations havebeen funded to focus on care transitions, applying lessons learned from clinical trials.Several have been notably successful in reducing readmissions within 30 days.36Given that studies have shown readmissions within 30-days to be related to quality ofcare, and that interventions have been able to reduce 30-day readmission rates, it isreasonable to consider an all-condition 30-day readmission rate as a quality measure.1.3 Approach to Measure DevelopmentWe developed this measure in consultation with national guidelines for publicly rep