2011 Measures Maintenance Technical Report: Acute Myocardial Infarction, Heart Failure, and Pneumonia 30 Day Risk Standardized Readmission Measures

Transcription

1 2011 Measures Maintenance Technical Report: Acute Myocardial Infarction, Heart Failure, and Pneumonia 30 Day Risk Standardized Readmission Measures Submitted By Yale New Haven Health Services Corporation / Center for Outcomes Research & Evaluation (YNHHSC/CORE): Susannah M. Bernheim, M.D, M.H.S Zhenqiu Lin, Ph.D. Jacqueline N. Grady, M.S. Kanchana R. Bhat, M.P.H. Haiyan Wang, M.D., M.S. Yongfei Wang, M.S.* Zameer Abedin, B.A. Mayur M. Desai, Ph.D., M.P.H.** Shu Xia Li, Ph.D. Smitha Vellanky, M.Sc Elizabeth E. Drye, M.D., S.M.* Harlan M. Krumholz, M.D., S.M.* *Yale School of Medicine **Yale School of Public Health Under Contract # HHSM I/HHSM 500 T0001, Modification No from the Centers for Medicare & Medicaid Services Prepared For: Centers for Medicare & Medicaid Services (CMS) April 13, 2011

2 TABLE OF CONTENTS LIST OF TABLES... 3 LIST OF FIGURES... 4 ACKNOWLEDGEMENTS INTRODUCTION Background on Readmission Measures Overview of Measure Methodology Goals of Measures Maintenance UPDATES TO METHODS Refinements to the Readmission Models Inclusion of VA Data into Models Inclusion of Additional Pneumonia Codes Updates to the CC Map Changes to SAS Analytic Package (SAS Pack) FINAL MODELS AND ASSESSMENT OF PERFORMANCE Overview of Methodology and Results AMI Readmission Model Index Cohort Admissions Not Counted As Readmissions Frequency of AMI Model Variables AMI Model Parameters and Performance Distribution of Hospital Volumes and RSRRs HF Readmission Model Index Cohort Frequency of HF Model Variables HF Model Parameters and Performance Distribution of Hospital Volumes and RSRRs Pneumonia Readmission Model Index Cohort Frequency of Pneumonia Model Variables Pneumonia Model Parameters and Performance Distribution of Hospital Volumes and RSRRs SAS PACKS Revision to SAS Packs QUALITY ASSURANCE (QA) QA for Input Data REFERENCES APPENDIX Readmission Measures Maintenance

3 LIST OF TABLES Table 1 Frequency of AMI Model Variables over Different Time Periods Table 2 Adjusted OR and 95% CIs for the AMI HGLM over Different Time Periods Table 3 AMI Generalized Linear Modeling (GLM) Performance over Different Time Periods Table 4 Distribution of Hospital AMI Volumes and RSRRs over Different Time Periods Table 5 Frequency of HF Model Variables over Different Time Periods Table 6 Adjusted OR and 95% CIs for the HF HGLM over Different Time Periods Table 7 HF GLM Performance over Different Time Periods Table 8 Distribution of Hospital HF Volumes and RSRRs over Different Time Periods Table 9 Frequency of Pneumonia Model Variables over Different Time Periods Table 10 Adjusted OR and 95% CIs for the Pneumonia HGLM over Different Time Periods Table 11 Pneumonia GLM Performance over Different Time Periods Table 12 Distribution of Hospital Pneumonia Volumes and RSRRs by Time Period Readmission Measures Maintenance

4 LIST OF FIGURES Figure 1 Admission Sample for AMI in the Calendar Year Dataset Figure 2 Distribution of 30 Day AMI RSRRs in the Calendar Year Dataset Figure 3 Admission Sample for HF in the Calendar Year Dataset Figure 4 Distribution of 30 Day HF RSRRs in the Calendar Year Dataset Figure 5 Admission Sample for Pneumonia in the Calendar Year Dataset Figure 6 Distribution of 30 Day Pneumonia RSRRs in the Calendar Year Dataset Figure 7 YNHHSC/CORE QA Processes Readmission Measures Maintenance

5 ACKNOWLEDGEMENTS This work is a collaborative effort, and the authors gratefully acknowledge and thank Mai Hubbard, Angela Merrill, Sandi Nelson, Eric Schone, and Marian Wrobel from Mathematica Policy Research, Inc.; Marta Render, Stacy Poe, and Ron Freyberg from the Veterans Health Administration; Sharon Lise Normand from Harvard Medical School, Department of Health Care Policy and Harvard School of Public Health, Department of Biostatistics; Jennifer Mattera and Lori Geary from Yale New Haven Health Services Corporation/Center for Outcomes Research and Evaluation; and Lein Han and Michael Rapp at the Centers for Medicare & Medicaid Services for their contributions to this work. Readmission Measures Maintenance

6 1. INTRODUCTION 1.1 Background on Readmission Measures In July 2009, the Centers for Medicare & Medicaid Services (CMS) began publicly reporting 30 day risk standardized readmission rates (RSRRs) for acute myocardial infarction (AMI), heart failure (HF), and pneumonia for the nation s non federal * acute care hospitals, including critical access hospitals. These three measures complement the 30 day mortality measures CMS reports for AMI, HF, and pneumonia. 1 2 The readmission measures are posted on the Hospital Compare Web site ( and CMS updates them annually. Until now, only non federal acute care hospitals have been included in public reporting. As discussed below, the 2011 updates for the readmission measures now include hospitalizations for patients admitted for AMI, HF, or pneumonia in the Veterans Health Administration (VA) hospitals. These measures were originally developed by a team of clinical and statistical experts from Yale New Haven Health Services Corporation/Center for Outcomes Research & Evaluation (YNHHSC/CORE), Yale University, and Harvard University. 3 5 All three measures are consistent with the American Heart Association standards for outcomes measures suitable for public reporting 6 and have been endorsed by the National Quality Forum (NQF). CMS contracted with YNHHSC/CORE to prepare the 30 day AMI, HF, and pneumonia readmission measures for 2011 public reporting through a process of measures maintenance. This report summarizes our measures maintenance activities, describes the updates made to the measures this year, and presents the 2011 models. The most significant update for 2011 reporting is the inclusion of VA hospitals in the measures. This report is a supplement to and an update of the prior methodology reports produced for each measure, rather than a comprehensive description of measure methods. The reports that present the measure methodology in full for each measure are available at QualityNet ( Hospital 30 Day Acute Myocardial Infarction Readmission Measure: Methodology (2008) 3 Hospital 30 Day Heart Failure Readmission Measure: Methodology (2008) 4 Hospital 30 Day Pneumonia Readmission Measure: Methodology (2008) Measure Maintenance Technical Report: Acute Myocardial Infarction, Heart Failure, and Pneumonia 30 Day Risk Standardized Readmission Measures Measures Maintenance Technical Report: Acute Myocardial Infarction, Heart Failure, and Pneumonia 30 Day Risk Standardized Readmission Measures 8 The AMI, HF, and pneumonia readmission measure methodologies are also described in the peer reviewed medical literature * Note: Includes Indian Health Services hospitals Readmission Measures Maintenance

7 1.2 Overview of Measure Methodology The 2011 risk adjusted readmission measures use the NQF endorsed methodology set forth in the initial measure methodology reports 3 5 (with slight refinements to the measures as described in the two readmission measures maintenance reports. 7,8 ) Below, we provide an overview of the methodology. The updates for 2011 are found in Section 2. The readmission measures use hierarchical generalized linear modeling (HGLM) to create an RSRR at the hospital level. The measures incorporate administrative claims data for each patient from one year prior to hospital admission to adjust for case mix differences at hospitals. Cohort Index Cohort (Included hospitalizations) The readmission measures include admissions for Medicare FFS and VA beneficiaries aged 65 years discharged from non federal acute care hospitals or VA hospitals having a principal discharge diagnosis of AMI, HF, or pneumonia. For specific International Classification of Diseases, 9th Revision, Clinical Modification (ICD 9 CM) codes used to define the inclusion cohort for each condition, refer to Sections 3.2.1, 3.3.1, and for AMI, HF, and pneumonia respectively. An index admission is the hospitalization considered for the readmission outcome determination. CMS FFS beneficiaries with an index hospitalization within a non federal hospital are included if they have been enrolled in Part A and Part B Medicare for the 12 months prior to the date of admission to ensure a full year of administrative data for risk adjustment. (This requirement is dropped for patients with an index admission within a VA hospital.) We restrict the index cohort to hospitalizations for patients age 65 years and older because national datasets on younger patients are not currently available. The CMS measures were developed using Medicare FFS administrative data but are designed for us in all payer claims datasets. Cohort Exclusions (Excluded Admissions) The measures exclude admissions for patients: with an in hospital death (because they are not eligible for readmission) without at least 30 days post discharge enrollment in FFS Medicare (because the 30 day readmission outcome cannot be assessed in this group). This exclusion applies only to patients who have index admissions in non VA hospitals. who were transferred to another acute care facility as described below (because we are focusing on discharges to non acute care settings) who were discharged against medical advice (AMA) (because providers did not have the opportunity to deliver full care and prepare the patient for discharge). In addition, if a patient has more than one admission within 30 days of discharge from the index hospitalization, only one is counted as a readmission, as we are interested in a dichotomous yes/no readmission outcome as opposed to the number of readmissions. No admissions within Readmission Measures Maintenance

8 30 days of discharge from an index admission are considered as additional index admissions. The next eligible admission after the 30 day time period following an index admission will be considered another index admission. The number of admissions excluded based on each criterion is available in Figures 1, 3, and 5 for AMI, HF, and pneumonia, respectively. An additional exclusion criterion for the AMI cohort is that we exclude patients discharged on the same day as their index admission. Transferred Patients For patients who are transferred between one acute care hospital and another, the measures consider these multiple contiguous hospitalizations as a single acute episode of care. Readmission for transferred patients is attributed to the hospital that ultimately discharges the patient to a non acute care setting (e.g., to home or a skilled nursing facility). Thus, for patients who are transferred between two or more hospitals, if the patient is readmitted in the 30 days following the final hospitalization, the readmission is attributed to the final hospital. Outcome 30 Day Timeframe The measures assess readmissions within a 30 day period from the date of discharge from an index hospitalization. This standard time period is necessary so that the outcome for each patient is measured consistently. Outcomes occurring within 30 days of discharge can be strongly influenced by hospital care and the early transition to the outpatient setting. The use of the 30 day timeframe is a clinically meaningful period for hospitals to collaborate with their communities in an effort to reduce readmissions. All Cause Readmission The measures include readmissions for all causes, regardless of the principal diagnosis of the readmission, because from a patient perspective, readmission from any cause is an adverse event. In addition, it is difficult to make inferences about quality issues and accountability based solely on the documented cause of readmission. For example, a patient with HF who develops a hospital acquired infection may ultimately be readmitted for sepsis. In this context it would be inappropriate to consider the readmission to be unrelated to the care the patient received for HF during the first hospitalization. The AMI measure, however, does not count planned readmissions for revascularization as part of the outcome because such re hospitalizations are conceptually a continuation of care for the index admission. See Section for details. Risk Adjustment Variables The measures adjust for key variables (e.g., demographic factors, comorbid diseases, and indicators of patient frailty) that are clinically relevant and have strong relationships with the Readmission Measures Maintenance

9 outcome. For each patient, covariates are obtained from Medicare administrative claims data and VA administrative data (for patients with a VA index admission) extending 12 months prior to and including the index admission. The measures seek to adjust for case mix differences based on the clinical status of the patient at the time of the index admission. Accordingly, only comorbidities that convey information about the patient at that time or in the 12 months prior and not complications that arise during the course of the hospitalization are included in the risk adjustment. The measures do not adjust for the patients admission source and their discharge disposition (e.g. skilled nursing facility) because these factors are associated with the structure of the health care system, not solely patients clinical risk factors. Regional differences in resource availability and practice patterns may exert an undue influence on model results. Moreover, the validity of these admission and discharge disposition codes is not known. The measures also do not adjust for socioeconomic status (SES) because the association between SES and health outcomes can be due, in part, to the differences in the quality of health care. Risk adjusting for patient SES would suggest that hospitals with low SES patients are held to different standards for the risk of readmission than hospitals treating higher SES patient populations. The intention is for the measures to adjust for patient demographic and clinical characteristics while illuminating important quality differences. This methodology is consistent with guidance from NQF. Refer to Tables 1, 5, and 9 in this report for the list of risk adjustment variables for AMI, HF, and pneumonia, respectively. Calculating the RSRR The measures estimate hospital level 30 day all cause RSRRs for each condition using HGLMs. In brief, the approach simultaneously models two levels (patient and hospital) to account for the variance in patient outcomes within and between hospitals. 12 At the patient level, it models the log odds of hospital readmission within 30 days of discharge using age, sex, selected clinical covariates, and a hospital specific intercept. At the hospital level, it models the hospital specific intercepts as arising from a normal distribution. The hospital intercept represents the underlying risk of a readmission at the hospital, after accounting for patient risk. The hospital specific intercepts are given a distribution in order to account for the clustering (non independence) of patients within the same hospital. 12 If there were no differences among hospitals, then after adjusting for patient risk, the hospital intercepts should be identical across all hospitals. The RSRR is calculated as the ratio of the number of adjusted actual readmissions (also referred to as predicted ) to the number of expected readmissions at a given hospital, multiplied by the national unadjusted readmission rate. For each hospital, the numerator of the ratio is the number of readmissions within 30 days predicted on the basis of the hospital s performance with its observed case mix, and the denominator is the number of readmissions expected on the basis of the nation s performance with that hospital s case mix. This approach is analogous to a ratio of observed to expected used in other types of statistical analyses. It conceptually allows for a comparison of a particular hospital s performance given its case mix to an average hospital s performance with the same case mix. Thus, a lower ratio indicates lowerthan expected readmission or better quality, and a higher ratio indicates higher than expected readmission or worse quality. Readmission Measures Maintenance

10 The adjusted actual readmissions (the numerator) is calculated by regressing the risk factors (see Tables 1, 5, and 9 for AMI, HF, and pneumonia, respectively) and the hospital specific intercept on the risk of readmission, multiplying the estimated regression coefficients by the patient characteristics in the hospital, transforming, and then summing over all patients attributed to the hospital to get a value. The expected number of readmissions (the denominator) is obtained by regressing the risk factors and a common intercept on the readmission outcome using all hospitals in our sample, multiplying the subsequent estimated regression coefficients by the patient characteristics observed in the hospital, transforming, and then summing over all patients in the hospital to get a value. To assess hospital performance in any reporting period, we re estimate the model coefficients using the years of data in that period. The statistical models used are described in full in the original methodology reports Goals of Measures Maintenance Measures maintenance is a process to continually improve the measures. Conducted annually, it is an opportunity to reflect on and respond to comments made in the last year of public reporting and to incorporate advances in the science, and any changes in coding. It ensures that the risk standardized readmission models are continually assessed and remain valid given possible changes in the data over time, and it allows for model refinements. As described in this report, for 2011 public reporting, we undertook the following measures maintenance activities: including VA hospitals into the readmission measures; incorporating ICD 9 CM coding updates; validating the performance of each condition specific model and its corresponding riskadjustment variables in three recent one year datasets (2007, 2008, and 2009); evaluating and validating model performance in the three year combined dataset ( ); and updating the Quality Assurance (QA) process and SAS pack documentation. Readmission Measures Maintenance

11 2. UPDATES TO METHODS 2.1 Refinements to the Readmission Models For 2011 public reporting, we made the following refinements to the model: incorporated VA data added codes to the pneumonia cohort incorporated the ICD 9 CM coding updates to the Condition Categories (CC) map We assessed the effects of these changes using admissions with discharge dates from January 1, 2007, through December 31, 2009 (this dataset is referred to in this report as the calendar year dataset) as well as using the admissions with discharge dates in the three calendar years of 2007, 2008, and 2009 separately. These changes are discussed in more detail below Inclusion of VA Data into Models Modification: In addition to admissions to US non federal acute care hospitals in the index cohorts of previous years, this year s index cohort for public reporting includes hospitalizations for patients admitted for AMI, HF or pneumonia within a VA hospital. Rationale: Inclusion of the VA hospitals allows for a more inclusive perspective of the relative quality of US hospitals. Effect on Patient Cohorts: For three year combined data ( ), adding VA hospitals increased the AMI cohort by 10,593; the HF cohort by 35,623; and the pneumonia cohort by 30,700. Approach to incorporating VA hospitals: Below we briefly describe the approach used to incorporate data from the VA into the measures and any changes made to prior measure methodologies to incorporate VA index admissions. Data Source: The data used to identify index admissions within the VA as well as inpatient and outpatient historical claims for patients with a VA index hospitalization come from the national VA database in Austin, TX. We used VA Medical SAS datasets, which include administrative data extracted from the National Patient Care Database, originally constituted from the patient treatment files of each VA hospital. Cohort: The index hospitalizations for patients hospitalized within the VA were selected based on a principal discharge diagnosis of AMI, HF, or pneumonia using the same ICD 9 CM codes and in the same way as the index hospitalizations for CMS beneficiaries in non federal hospitals were selected. (See sections 3.2.1, 3.3.1, and for AMI, HF, and pneumonia, respectively). Linking to CMS data: Using social security number, date of birth, and sex, each patient with an index hospitalization within the VA was linked to the CMS enrollment file to identify their corresponding CMS HIC number (97% linked). This linking facilitates Readmission Measures Maintenance

12 obtaining CMS administrative claims data for patients with index admissions in VA hospitals. Index admissions which occurred in either VA or non federal facilities (for CMS beneficiaries) were combined into a single cohort. This combined cohort of VA and nonfederal index hospitalizations allowed for identification of patients transferred between VA and non federal hospitals. Also, for all index admissions (in VA and non federal hospitals), CMS administrative claims data were obtained for inpatient stays and outpatient visits in the 12 months prior to the index admission for risk adjustment (consistent with prior methodology for the CMS only cohort) and for inpatient stays in the 30 days following discharge. Thus, for any patient included in the measures based on an index hospitalization at a VA hospital, we obtained both VA and CMS administrative data for use in risk adjustment and readmission identification. Also, for index hospitalizations at non federal hospitals, the measures may capture a readmission to a VA hospital (for patients with any index admission to the VA). Inclusion/Exclusion criteria: The inclusion and exclusion criteria for the VA index hospitalizations were the same as those for index hospitalizations within non federal hospitals (for CMS beneficiaries), except that those patients with an index hospitalization within the VA were not required to have a full year enrollment in CMS prior to hospitalization for inclusion. Having 30 days post discharge enrollment information is a requirement for Medicare beneficiaries in the measure, but this criterion was also waived for the VA beneficiaries. Finally, all stays of 24 hours or less are considered observation status within the VA and thus no such stays were included as index hospitalizations within VA hospitals. Risk Adjustment: The VA administrative data includes 34 diagnosis and 46 procedure codes (as opposed to 10 and 6, respectively, in CMS administrative data). All diagnosis codes were retained and the first 6 procedure codes were retained for the index hospitalizations. All diagnosis and procedure codes were retained for visits prior to the index hospitalization for use in risk adjustment. Hospital Identifiers: VA hospitals (stations) that share a CMS Certification Number (CCN) were treated as single institutions to be consistent with the approach used for other VA quality measures. Vital status and Demographics: For index admissions within the VA system, data on demographics and vital status came from the VA data. However, if there was no date of death in the VA data, but there was a date of death in the CMS EDB, this information was used. Similar considerations were used for date of birth and gender. Model Validation: The measure methodologies developed in the Medicare FFS populations were used without modification on the combined Medicare and VA populations. We did, however, validate the measures for VA hospitals. To evaluate whether the measures, as currently constructed, perform well in the VA population we completed a medical record measure validation. We created a medical record model for VA patients, applied the administrative model to same set of VA patients, and then compared the performance of VA hospitals based on the medical record model versus Readmission Measures Maintenance

13 hospital performance based on the administrative model. We found a very high degree of correlation between the hospital RSRRs estimated by the two measures. Detailed results can be obtained upon request to CMS Inclusion of Additional Pneumonia Codes Modification: Inclusion of the following codes in the pneumonia measure: : Methicillin resistant pneumonia due to Staphylococcus aureus : Influenza due to identified novel H1N1 influenza virus with pneumonia Rationale: The pneumonia measure, as developed and subsequently implemented, included pneumonias due to Staphylococcus aureus (482.41). Upon review of coding changes made in recent years, CMS is revising the pneumonia measure cohort definition to include cases of Methicillin resistant pneumonia due to Staphylococcus aureus (482.42). Previously, these patients were captured by the pneumonia measure under pneumonia due to Staphylococcus aureus (482.41), but the updated International Classification of Diseases, 9th revision, divides pneumonias due to Staphylococcus aureus into two categories: Methicillin susceptible (482.41) and Methicillin resistant (482.42). Similarly, the pneumonia measure includes viral pneumonia cases, including influenza. The latest version of the International Classification of Diseases, 9th revision, has added more specific codes to reflect the emergence of influenza due to novel H1N1 influenza virus. As such, CMS is revising the pneumonia measure cohort definition to add the code for Influenza due to identified novel H1N1 influenza virus with pneumonia (488.11). The cases represented by these new codes are consistent with the pneumonia measure cohort as previously defined, and are representative of the spectrum of care delivered to pneumonia patients. Effects on Patient Cohort: Overall, relatively few cases were coded as (N=11,137); no additional cases were captured by the code since it was not actively used before Updates to the CC Map Modification: A second CMS contractor, RTI International, updated the map linking ICD 9 CM codes to CCs clinically related groups of conditions used for measure riskadjustment variables to reflect ICD 9 CM codes in use during the full reporting period. RTI International, contracted by CMS to maintain the CC system, assigned new ICD 9 CM codes to the existing CCs based on their clinical expertise and the historical assignment of related ICD 9 CM codes to the CCs. Rationale: CMS revises the ICD 9 CM CC map annually to reflect changes in ICD 9 CM codes so that the measures will capture all relevant comorbidities coded in patient claims data. Readmission Measures Maintenance

14 Effects on Model Variables: The assignment of new codes and the removal of retired codes had little impact on the model variables since RTI assigned the majority of new codes, which were more specific versions of retired codes, to the same CCs as retired codes. For more details on the CC changes, see the Appendix for RTI s memo to CMS detailing the map changes. 2.2 Changes to SAS Analytic Package (SAS Pack) We revised the SAS pack to reflect all changes to the admission cohorts and models as needed, including any ad hoc patches to address data issues. The primary changes this year were made to include the increased number of diagnosis and procedure codes from the VA data. Also a change was made to incorporate the new pneumonia codes in the pneumonia specific SAS pack. The new SAS packs are named AMI_readmission_v2011, HF_readmission_v2011, and Pneumonia_readmission_v2011 (see Section 4 for details). Readmission Measures Maintenance

15 3. FINAL MODELS AND ASSESSMENT OF PERFORMANCE 3.1 Overview of Methodology and Results The readmission measures estimate hospital specific, 30 day all cause RSRRs using HGLMs. To adjust for differences in hospital case mix, the models control for patient risk factors, including age and comorbidities present at the time of admission. A brief description of the measure methodology and model risk adjustment variables is in Section 1.2 of this report, and further 3 5, 7, 8 details are available in prior technical reports. The measures link admissions for patients who are transferred between acute care hospitals into a single acute episode of care. The outcome for the patient is assigned to the last hospital in the sequence of transfers for the purposes of evaluating 30 day readmission. To evaluate the performance of the models used for 2011 reporting, we fit the revised models to three single, calendar year datasets (2007, 2008, and 2009) and to the combined three year calendar year dataset. We examined trends in the frequency of patient risk factors and model variable coefficients, and compared the model performance in each of these datasets. As indicated above in Section 2, the main update this year was to incorporate VA hospitals into the measures. We otherwise preserved the original methodology and did not, for example, reselect variables for inclusion into the models. For each of the three measures, we assessed HGLM performance in terms of discriminant ability and overall fit for each calendar year of data (2007, 2008, 2009) and for the three year combined period ( ). We computed two summary statistics for assessing model performance: the predictive ability and the area under the receiver operating characteristic (ROC) curve (c statistic), which is an indicator of the model s discriminant ability or ability to correctly classify those who are and are not readmitted within 30 days of discharge (potential values range from 0.5 meaning no better than chance to 1.0 meaning perfect discrimination). The data sources for these measures maintenance analyses are Medicare administrative claims and VA administrative data and enrollment information for hospitalizations which occurred in calendar years The datasets also contain associated inpatient, outpatient, and physician carriers administrative data for the prior 12 months and one month subsequent to the index hospitalization for patients admitted in each of these years. Please see the methodology reports 3 5 for further descriptions of these data sources as well as the updates from Section above for a description of equivalent VA data sources. The results of these analyses for each of the three measures (AMI, HF, and pneumonia) are presented below in Sections 3.2, 3.3, and 3.4, respectively. We also assessed model performance for each measure using preliminary public reporting data for 2011 (admissions with discharges between July 1, 2007, and June 30, 2010). The results (data not shown) were substantively similar to those for the calendar year dataset. Readmission Measures Maintenance

16 AMI Readmission Model Index Cohort The cohort includes admissions for Medicare FFS beneficiaries and VA beneficiaries age 65 years discharged from non federal acute care hospitals or VA hospitals having a principal discharge diagnosis of AMI (ICD 9 CM code 410.xx, excluding those with 410.x2 [AMI, subsequent episode of care]). The exclusion criteria for the measures are presented in Section 1.2 and the percentage of AMI patients meeting each exclusion criterion in the calendar year dataset is presented in Figure Admissions Not Counted As Readmissions Some AMI patients have planned readmissions for revascularization procedures for example, to perform percutaneous transluminal coronary angioplasty (PTCA) on a second vessel or a second location in the same vessel, or to perform coronary artery bypass graft (CABG) surgery after AMI and a period of recovery outside the hospital. Because admissions for PTCA or CABG may be staged or scheduled readmissions, we do not count as readmissions those admissions after discharge that include PTCA or CABG procedures unless the principal discharge diagnosis for the readmission is one of the following acute diagnoses, which are not consistent with a scheduled readmission: HF, AMI, unstable angina, arrhythmia, and cardiac arrest (i.e., readmissions with these diagnoses and a PTCA or CABG procedure are counted as readmissions). The ICD 9 CM procedure codes associated with PTCA and CABG revascularization procedures are: PTCA: 00.66, 36.06, CABG: The ICD 9 CM diagnosis codes associated with HF, AMI, unstable angina, arrhythmia, and cardiac arrest are: HF: , , , , , , , , , 428.xx AMI: 410.xx, except 410.x2 (AMI, subsequent episode of care) Unstable angina: 411.xx Arrhythmia: 427.xx, except Cardiac arrest: Readmission Measures Maintenance

17 Figure 1 Admission Sample for AMI in the Calendar Year Dataset Frequency of AMI Model Variables We examined the temporal variation in both overall readmission and frequency of clinical and demographic variables. Between 2007 and 2009, the crude readmission rate remained stable at approximately 20%. During this time period, although the frequency of most of the model variables remained relatively constant, there was a slight increase in the frequency of iron Readmission Measures Maintenance

18 deficiency/anemias/blood disease and a slight decrease in the frequency of valvular / rheumatic heart disease (Table 1) AMI Model Parameters and Performance Table 2 shows the risk adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for the AMI readmission model by individual year and for the combined calendar year dataset. Overall, the variable effect sizes were relatively constant across years. In addition, model performance was stable over the three year time period; the area under the ROC curve (c statistic) remained constant at 0.63 (Table 3) Distribution of Hospital Volumes and RSRRs Table 4 shows the distributions of hospital volumes and hospital RSRRs, as well as the between hospital variance, by individual year and for the combined calendar year dataset. Between 2007 and 2009, mean AMI volume decreased from 45.4 to 43.3 admissions per hospital. The mean RSRR was stable across the three year time period. The mean hospital RSRR in the combined three year dataset was 19.9% (range: 15.3% 26.8%). Between hospital variance win the combined dataset was (SE: 0.002). If there were no systematic differences between hospitals, the between hospital variance would be 0. Figure 2 shows the overall distribution of the hospital RSRRs for the combined calendar year dataset. The odds of all cause readmission if treated at a hospital one standard deviation above the national average were 1.34 times higher than the odds of all cause readmission if treated at a hospital one standard deviation below the national average. If there were no systematic differences between hospitals, the OR would be Readmission Measures Maintenance

19 Table 1 Frequency of AMI Model Variables over Different Time Periods Variable Total N 192, , , ,427 Crude readmission rate (%) Demographic Mean Age-65 (SD) 13.9 (8.0) 14.1 (8.1) 14.0 (8.2) 14.0 (8.1) Male (%) Cardiovascular (%) History of PTCA History of CABG Congestive heart failure (CC 80) Acute coronary syndrome (CC 81-82) Anterior myocardial infarction Other location myocardial infarction Angina pectoris, old MI (CC 83) Coronary atherosclerosis (CC 84) Valvular or rheumatic heart disease (CC 86) Specified arrhythmias (CC 92-93) Comorbidity (%) History of infection (CC 1, 3-6) Metastatic cancer or acute leukemia (CC 7) Cancer (CC 8-12) Diabetes mellitus (DM) or DM complications (CC 15-20, ) Protein-calorie malnutrition (CC 21) Disorders of fluid, electrolyte, acid-base (CC 22-23) Iron deficiency or other anemias and blood disease (CC 47) Dementia or other specified brain disorders (CC 49-50) Hemiplegia, paraplegia, paralysis, functional disability (CC 67-69, , ) Stroke (CC 95-96) Cerebrovascular disease (CC 97-99, 103) Vascular or circulatory disease (CC ) Chronic obstructive pulmonary disease (CC 108) Asthma (CC 110) Pneumonia (CC ) End stage renal disease or dialysis (CC ) Renal failure (CC 131) Other urinary tract disorders (CC 136) Decubitus ulcer or chronic skin ulcer (CC ) Readmission Measures Maintenance

20 Variable Table 2 Adjusted OR and 95% CIs for the AMI HGLM over Different Time Periods 2007 OR 2007 (95% CI) 2008 OR 2008 (95% CI) 2009 OR 2009 (95% CI) OR (95% CI) Demographic Age-65 (years above 65, continuous) 1.01 ( ) 1.01 ( ) 1.01 ( ) 1.01 ( ) Male 0.93 ( ) 0.95 ( ) 0.93 ( ) 0.94 ( ) Cardiovascular History of PTCA 0.91 ( ) 0.87 ( ) 0.84 ( ) 0.88 ( ) History of CABG 0.93 ( ) 0.92 ( ) 0.96 ( ) 0.93 ( ) Congestive heart failure (CC 80) 1.21 ( ) 1.23 ( ) 1.23 ( ) 1.22 ( ) Acute coronary syndrome (CC 81-82) 1.02 ( ) 1.04 ( ) 1.02 ( ) 1.03 ( ) Anterior myocardial infarction 1.16 ( ) 1.17 ( ) 1.20 ( ) 1.18 ( ) Other location myocardial infarction 0.88 ( ) 0.93 ( ) 0.94 ( ) 0.92 ( ) Angina pectoris, old MI (CC 83) 1.04 ( ) 1.02 ( ) 1.02 ( ) 1.03 ( ) Coronary atherosclerosis (CC 84) 0.90 ( ) 0.92 ( ) 0.92 ( ) 0.92 ( ) Valvular or rheumatic heart disease (CC 86) 1.14 ( ) 1.11 ( ) 1.13 ( ) 1.13 ( ) Specified arrhythmias (CC 92-93) 1.10 ( ) 1.10 ( ) 1.09 ( ) 1.10 ( ) Comorbidity History of infection (CC 1, 3-6) 1.05 ( ) 1.03 ( ) 1.06 ( ) 1.04 ( ) Metastatic cancer or acute leukemia (CC 7) 1.18 ( ) 1.20 ( ) 1.24 ( ) 1.21 ( ) Cancer (CC 8-12) 1.02 ( ) 1.01 ( ) 1.05 ( ) 1.03 ( ) Diabetes mellitus (DM) or DM complications (CC 15-20, ( ) 1.21 ( ) 1.20 ( ) 1.20 ( ) 120) Protein-calorie malnutrition (CC 21) 1.10 ( ) 1.14 ( ) 1.15 ( ) 1.13 ( ) Disorders of fluid, electrolyte, acid-base (CC 22-23) 1.10 ( ) 1.11 ( ) 1.13 ( ) 1.11 ( ) Iron deficiency or other anemias and blood disease (CC 47) 1.19 ( ) 1.18 ( ) 1.18 ( ) 1.18 ( ) Dementia or other specified brain disorders (CC 49-50) 1.02 ( ) 0.96 ( ) 0.96 ( ) 0.98 ( ) Hemiplegia, paraplegia, paralysis, functional disability (CC 67-69, 1.09 ( ) 1.08 ( ) 1.07 ( ) 1.08 ( ) , ) Stroke (CC 95-96) 1.02 ( ) 1.03 ( ) 1.05 ( ) 1.03 ( ) Cerebrovascular disease (CC 97-99, 103) 1.06 ( ) 1.06 ( ) 1.04 ( ) 1.05 ( ) Vascular or circulatory disease (CC ) 1.12 ( ) 1.10 ( ) 1.10 ( ) 1.11 ( ) Chronic obstructive pulmonary disease (CC 108) 1.26 ( ) 1.22 ( ) 1.26 ( ) 1.24 ( ) Asthma (CC 110) 1.00 ( ) 1.04 ( ) 1.01 ( ) 1.02 ( ) Pneumonia (CC ) 1.13 ( ) 1.15 ( ) 1.17 ( ) 1.15 ( ) End stage renal disease or dialysis (CC ) 1.34 ( ) 1.35 ( ) 1.37 ( ) 1.35 ( ) Renal failure (CC 131) 1.13 ( ) 1.19 ( ) 1.19 ( ) 1.17 ( ) Other urinary tract disorders (CC 136) 1.08 ( ) 1.08 ( ) 1.07 ( ) 1.08 ( ) Decubitus ulcer or chronic skin ulcer (CC ) 1.05 ( ) 1.13 ( ) 1.09 ( ) 1.09 ( ) Between Hospital Variance (SE) (0.003) (0.003) (0.003) (0.002) Readmission Measures Maintenance

21 Table 3 AMI Generalized Linear Modeling (GLM) Performance over Different Time Periods Characteristic c-statistic Predictive ability, % (lowest decile highest decile) Readmission Measures Maintenance

22 Table 4 Distribution of Hospital AMI Volumes and RSRRs over Different Time Periods Characteristic Number of Hospitals 4,236 4,210 4,119 4,576 Hospital Volume Mean (SD) 45.4 (71.6) 44.9 (69.5) 43.3 (66.2) (200.3) Range (min. max.) , th percentile th percentile th percentile RSRR (%) Mean (SD) 20.0 (0.7) 19.9 (0.6) 19.8 (0.9) 19.9 (1.0) Range (min. max.) th percentile th percentile th percentile Readmission Measures Maintenance

23 Figure 2 Distribution of 30 Day AMI RSRRs in the Calendar Year Dataset N= 4,576 hospitals Readmission Measures Maintenance

24 HF Readmission Model Index Cohort The cohort includes admissions for Medicare FFS beneficiaries and VA beneficiaries age 65 years discharged from non federal acute care hospitals or VA hospitals having a principal discharge diagnosis of HF (ICD 9 CM codes , , , , , , , , , and 428.xx). The exclusion criteria for the measures are presented in Section 1.2, and the percentage of HF patients meeting each exclusion criterion in the calendar year dataset is presented in Figure 3. Figure 3 Admission Sample for HF in the Calendar Year Dataset Readmission Measures Maintenance

25 3.3.2 Frequency of HF Model Variables We examined the temporal variation in both overall readmission and frequency of clinical and demographic variables. Between 2007 and 2009, the crude readmission rate remained stable at approximately 25%. During this time period, although the frequency of most of the model variables remained relatively constant, there was a slight increase in the frequency of cardiorespiratory failure and shock, protein calorie malnutrition, iron deficiency/anemias/ blood disease, and renal failure, and a decrease in the frequency of history of CABG (Table 5) HF Model Parameters and Performance Table 6 shows the risk adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for the HF readmission model by individual year and for the combined calendar year dataset. Overall, the variable effect sizes were relatively constant across years. In addition, model performance was stable over the three year time period; the area under the ROC curve (c statistic) remained constant at 0.60 across years. (Table 7) Distribution of Hospital Volumes and RSRRs Table 8 shows the distributions of hospital volumes and hospital RSRRs, as well as the between hospital variance, by individual year and for the combined calendar year dataset. Between 2007 and 2009, mean HF volume decreased from 98.7 to 94.8 admissions per hospital. The mean RSRR remained stable across the three year time period. The mean hospital RSRR in the combined three year dataset was 24.8% (range: 17.0% 33.0%). Between hospital variance in the combined dataset was (SE: 0.001). If there were no systematic differences between hospitals, the between hospital variance would be 0. Figure 4 shows the overall distribution of the RSRRs for the combined calendar year dataset. The odds of all cause readmission if treated at a hospital one standard deviation above the national average were 1.36 times higher than the odds of all cause readmission if treated at a hospital one standard deviation below the national average. If there were no systematic differences between hospitals, the OR would be Readmission Measures Maintenance

26 Table 5 Frequency of HF Model Variables over Different Time Periods Variable Total N 470, , ,253 1,365,780 Crude readmission rate (%) Demographic Mean Age-65 (SD) 15.4 (7.9) 15.6 (8.0) 15.7 (8.1) 15.6 (8.0) Male (%) Cardiovascular (%) History of CABG Cardio-respiratory failure or shock (CC 79) Congestive heart failure (CC 80) Acute coronary syndrome (CC 81-82) Coronary atherosclerosis or angina (CC 83-84) Valvular or rheumatic heart disease (CC 86) Specified arrhythmias (CC 92-93) Other or unspecified heart disease (CC 94) Vascular or circulatory disease (CC ) Comorbidity (%) Metastatic cancer or acute leukemia (CC 7) Cancer (CC 8-12) Diabetes or DM complications (CC 15-20, ) Protein-calorie malnutrition (CC 21) Disorders of fluid, electrolyte, acid-base (CC 22-23) Liver or biliary disease (CC 25-30) Peptic ulcer, hemorrhage, other specified gastrointestinal disorders (CC 34) Other gastrointestinal disorders (CC 36) Severe hematological disorders (CC 44) Iron deficiency or other anemias and blood disease (CC 47) Dementia or other specified brain disorders (CC 49-50) Drug/alcohol abuse/dependence/psychosis (CC 51-53) Major psychiatric disorders (CC 54-56) Depression (CC 58) Other psychiatric disorders (CC 60) Hemiplegia, paraplegia, paralysis, functional disability (CC 67-69, , ) Stroke (CC 95-96) Chronic obstructive pulmonary disease (CC 108) Fibrosis of lung or other chronic lung disorders (CC 109) Readmission Measures Maintenance

27 Variable Asthma (CC 110) Pneumonia (CC ) End stage renal disease or dialysis (CC ) Renal failure (CC 131) Nephritis (CC 132) Other urinary tract disorders (CC 136) Decubitus ulcer or chronic skin ulcer (CC ) Readmission Measures Maintenance

28 Variable Table 6 Adjusted OR and 95% CIs for the HF HGLM over Different Time Periods 2007 OR 2007 (95% CI) 2008 OR 2008 (95% CI) 2009 OR 2009 (95% CI) OR (95% CI) Demographic Age-65 (years above 65, continuous) 1.00 ( ) 1.00 ( ) 1.00 ( ) 1.00 ( ) Male 1.01 ( ) 1.01 ( ) 1.01 ( ) 1.01 ( ) Cardiovascular History of CABG 0.91 ( ) 0.90 ( ) 0.91 ( ) 0.90 ( ) Cardio-respiratory failure or shock (CC 79) 1.09 ( ) 1.11 ( ) 1.11 ( ) 1.11 ( ) Congestive heart failure (CC 80) 1.07 ( ) 1.09 ( ) 1.11 ( ) 1.09 ( ) Acute coronary syndrome (CC 81-82) 1.11 ( ) 1.13 ( ) 1.11 ( ) 1.12 ( ) Coronary atherosclerosis or angina (CC 83-84) 1.09 ( ) 1.07 ( ) 1.07 ( ) 1.07 ( ) Valvular or rheumatic heart disease (CC 86) 1.09 ( ) 1.07 ( ) 1.07 ( ) 1.08 ( ) Specified arrhythmias (CC 92-93) 1.06 ( ) 1.07 ( ) 1.08 ( ) 1.07 ( ) Other or unspecified heart disease (CC 94) 1.05 ( ) 1.05 ( ) 1.04 ( ) 1.05 ( ) Vascular or circulatory disease (CC ) 1.08 ( ) 1.07 ( ) 1.06 ( ) 1.07 ( ) Comorbidity Metastatic cancer or acute leukemia (CC 7) 1.13 ( ) 1.16 ( ) 1.13 ( ) 1.14 ( ) Cancer (CC 8-12) 1.01 ( ) 1.02 ( ) 1.00 ( ) 1.01 ( ) Diabetes or DM complications (CC 15-20, ) 1.08 ( ) 1.07 ( ) 1.08 ( ) 1.08 ( ) Protein-calorie malnutrition (CC 21) 1.12 ( ) 1.12 ( ) 1.10 ( ) 1.11 ( ) Disorders of fluid, electrolyte, acid-base (CC 22-23) 1.11 ( ) 1.13 ( ) 1.12 ( ) 1.12 ( ) Liver or biliary disease (CC 25-30) 1.08 ( ) 1.07 ( ) 1.08 ( ) 1.07 ( ) Peptic ulcer, hemorrhage, other specified gastrointestinal disorders 1.06 ( ) 1.07 ( ) 1.07 ( ) 1.06 ( ) (CC 34) Other gastrointestinal disorders (CC 36) 1.06 ( ) 1.04 ( ) 1.06 ( ) 1.05 ( ) Severe hematological disorders (CC 44) 1.14 ( ) 1.15 ( ) 1.14 ( ) 1.14 ( ) Iron deficiency or other anemias and blood disease (CC 47) 1.08 ( ) 1.08 ( ) 1.08 ( ) 1.08 ( ) Dementia or other specified brain disorders (CC 49-50) 1.01 ( ) 1.02 ( ) 1.00 ( ) 1.01 ( ) Drug/alcohol abuse/dependence/psychosis (CC 51-53) 1.10 ( ) 1.10 ( ) 1.12 ( ) 1.11 ( ) Major psychiatric disorders (CC 54-56) 1.03 ( ) 1.05 ( ) 1.02 ( ) 1.03 ( ) Depression (CC 58) 1.01 ( ) 1.01 ( ) 1.01 ( ) 1.01 ( ) Other psychiatric disorders (CC 60) 1.09 ( ) 1.06 ( ) 1.07 ( ) 1.07 ( ) Hemiplegia, paraplegia, paralysis, functional disability (CC 67-69, 1.03 ( ) 1.04 ( ) 1.03 ( ) 1.03 ( ) , ) Stroke (CC 95-96) 1.03 ( ) 1.03 ( ) 1.02 ( ) 1.02 ( ) Chronic obstructive pulmonary disease (CC 108) 1.13 ( ) 1.15 ( ) 1.13 ( ) 1.13 ( ) Fibrosis of lung or other chronic lung disorders (CC 109) 1.07 ( ) 1.05 ( ) 1.06 ( ) 1.06 ( ) Asthma (CC 110) 1.03 ( ) 1.03 ( ) 1.02 ( ) 1.02 ( ) Readmission Measures Maintenance

29 Variable OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) Pneumonia (CC ) 1.09 ( ) 1.09 ( ) 1.10 ( ) 1.09 ( ) End stage renal disease or dialysis (CC ) 1.17 ( ) 1.16 ( ) 1.15 ( ) 1.16 ( ) Renal failure (CC 131) 1.20 ( ) 1.18 ( ) 1.19 ( ) 1.19 ( ) Nephritis (CC 132) 1.10 ( ) 1.13 ( ) 1.08 ( ) 1.10 ( ) Other urinary tract disorders (CC 136) 1.08 ( ) 1.06 ( ) 1.07 ( ) 1.07 ( ) Decubitus ulcer or chronic skin ulcer (CC ) 1.12 ( ) 1.10 ( ) 1.10 ( ) 1.10 ( ) Between Hospital Variance (SE) (0.002) (0.002) (0.002) (0.001) Readmission Measures Maintenance

30 Table 7 HF GLM Performance over Different Time Periods Characteristic c-statistic Predictive ability, % (lowest decile highest decile) Readmission Measures Maintenance

31 Table 8 Distribution of Hospital HF Volumes and RSRRs over Different Time Periods Characteristic Number of Hospitals 4,767 4,748 4,728 4,882 Hospital Volume Mean (SD) 98.7 (114.1) 94.2 (109.3) 94.8 (112.6) (331.9) Range (min. max.) 1 1, , , , th percentile th percentile th percentile RSRR (%) Mean (SD) 24.7 (1.4) 24.9 (1.4) 24.8 (1.2) 24.8 (1.8) Range (min. max.) th percentile th percentile th percentile Readmission Measures Maintenance

32 Figure 4 Distribution of 30 Day HF RSRRs in the Calendar Year Dataset N= 4,882 hospitals Readmission Measures Maintenance

33 Pneumonia Readmission Model Index Cohort The cohort includes admissions for Medicare FFS beneficiaries and VA beneficiaries age 65 years discharged from non federal acute care hospitals or VA hospitals having a principal discharge diagnosis of pneumonia (ICD 9 CM codes 480.0, 480.1, 480.2, 480.3, 480.8, 480.9, 481, 482.0, 482.1, 482.2, , , , , , , , , , , , , , 482.9, 483.0, 483.1, 483.8, 485, 486, 487.0, and ). The exclusion criteria for the measures are presented in section 1.2, and the percentage of pneumonia patients meeting each exlcusion criterion in the calendar year dataset is presented in Figure 5. Readmission Measures Maintenance

34 Figure 5 Admission Sample for Pneumonia in the Calendar Year Dataset Frequency of Pneumonia Model Variables We examined the temporal variation in both overall readmission and frequency of clinical and demographic variables. Between 2007 and 2009, the crude readmission rate remained stable across years at just over 18% (Table 9). During this time period, although the frequency of most of the model variables remained relatively constant, there was a slight increase in the frequency of diabetes mellitus or DM complications, protein calorie malnutrition, iron deficiency/anemias/ blood disease, other lung diseases and renal failure, and a decrease in the frequency of COPD (Table 9). Readmission Measures Maintenance

35 3.4.3 Pneumonia Model Parameters and Performance Table 10 shows the risk adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for the pneumonia readmission model by individual year and for the combined calendar year dataset. Overall, the variable effect sizes were relatively constant across years. In addition, model performance was stable over the three year time period; the area under the ROC curve (c statistic) remained constant at 0.63 across years Distribution of Hospital Volumes and RSRRs Table 12 shows the distributions of hospital volumes and hospital RSRRs, as well as the between hospital variance, by individual year and for the combined calendar year dataset. Between 2007 and 2009, mean pneumonia volume decreased from 87.5 to 78.2 admissions per hospital. The mean RSRR was stable over the three year period. The mean hospital RSRR in the combined three year dataset was 18.4% (range: 13.8% 26.4%). Between hospital variance in the combined dataset was (SE: 0.001). If there were no systematic differences between hospitals, the between hospital variance would be 0. Figure 6 shows the overall distribution of the RSRR for the combined calendar year dataset. The odds of all cause readmission if treated at a hospital one standard deviation above the national average were 1.39 times higher than the odds of all cause readmission if treated at a hospital one standard deviation below the national average. If there were no systematic differences between hospitals, the OR would be Readmission Measures Maintenance

36 Table 9 Frequency of Pneumonia Model Variables over Different Time Periods Variable Total N 422, , ,212 1,199,324 Crude readmission rate (%) Demographic Mean Age-65 (SD) 15.0 (8.0) 15.3 (8.1) 15.1 (8.2) 15.1 (8.1) Male (%) Comorbidity (%) History of CABG History of infection (CC 1, 3-6) Septicemia/shock (CC 2) Metastatic cancer or acute leukemia (CC 7) Lung or other severe cancers (CC 8) Other major cancers (CC 9-10) Diabetes mellitus (DM) or DM complications (CC 15-20, ) Protein-calorie malnutrition (CC 21) Disorders of fluid, electrolyte, acid-base (CC 22-23) Other gastrointestinal disorders (CC 36) Severe hematological disorders (CC 44) Iron deficiency or other anemias and blood disease (CC 47) Dementia or other specified brain disorders (CC 49-50) Drug/alcohol abuse/dependence/psychosis (CC 51-53) Major psychiatric disorders (CC 54-56) Other psychiatric disorders (CC 60) Hemiplegia, paraplegia, paralysis, functional disability (CC 67-69, , ) Cardio-respiratory failure or shock (CC 79) Congestive heart failure (CC 80) Acute coronary syndrome (CC 81-82) Coronary atherosclerosis or angina (CC 83-84) Valvular or rheumatic heart disease (CC 86) Specified arrhythmias (CC 92-93) Stroke (CC 95-96) Vascular or circulatory disease (CC ) Chronic obstructive pulmonary disease (CC 108) Fibrosis of lung or other chronic lung disorders (CC 109) Asthma (CC 110) Readmission Measures Maintenance

37 Variable Pneumonia (CC ) Pleural effusion/pneumothorax (CC 114) Other lung disorders (CC 115) End stage renal disease or dialysis (CC ) Renal failure (CC 131) Urinary tract infection (CC 135) Other urinary tract disorders (CC 136) Decubitus ulcer or chronic skin ulcer (CC ) Vertebral fractures (CC 157) Other injuries (CC 162) Readmission Measures Maintenance

38 Variable Table 10 Adjusted OR and 95% CIs for the Pneumonia HGLM over Different Time Periods 2007 OR 2007 (95% CI) 2008 OR 2008 (95% CI) 2009 OR 2009 (95% CI) OR (95% CI) Demographic Age-65 (years above 65, continuous) 1.00 ( ) 1.00 ( ) 1.00 ( ) 1.00 ( ) Male 1.09 ( ) 1.08 ( ) 1.06 ( ) 1.08 ( ) Comorbidity History of CABG 0.88 ( ) 0.90 ( ) 0.89 ( ) 0.89 ( ) History of infection (CC 1, 3-6) 1.05 ( ) 1.06 ( ) 1.04 ( ) 1.05 ( ) Septicemia/shock (CC 2) 1.07 ( ) 1.08 ( ) 1.07 ( ) 1.07 ( ) Metastatic cancer or acute leukemia (CC 7) 1.21 ( ) 1.21 ( ) 1.20 ( ) 1.21 ( ) Lung or other severe cancers (CC 8) 1.22 ( ) 1.22 ( ) 1.19 ( ) 1.21 ( ) Other major cancers (CC 9-10) 1.05 ( ) 1.02 ( ) 1.04 ( ) 1.04 ( ) Diabetes mellitus (DM) or DM complications (CC 15-20, ) 1.08 ( ) 1.09 ( ) 1.08 ( ) 1.08 ( ) Protein-calorie malnutrition (CC 21) 1.14 ( ) 1.17 ( ) 1.17 ( ) 1.16 ( ) Disorders of fluid, electrolyte, acid-base (CC 22-23) 1.15 ( ) 1.16 ( ) 1.17 ( ) 1.16 ( ) Other gastrointestinal disorders (CC 36) 1.05 ( ) 1.04 ( ) 1.03 ( ) 1.04 ( ) Severe hematological disorders (CC 44) 1.22 ( ) 1.23 ( ) 1.19 ( ) 1.21 ( ) Iron deficiency or other anemias and blood disease (CC 47) 1.12 ( ) 1.13 ( ) 1.12 ( ) 1.13 ( ) Dementia or other specified brain disorders (CC 49-50) 1.05 ( ) 1.04 ( ) 1.00 ( ) 1.02 ( ) Drug/alcohol abuse/dependence/psychosis (CC 51-53) 1.11 ( ) 1.10 ( ) 1.09 ( ) 1.10 ( ) Major psychiatric disorders (CC 54-56) 1.04 ( ) 1.06 ( ) 1.04 ( ) 1.04 ( ) Other psychiatric disorders (CC 60) 1.12 ( ) 1.09 ( ) 1.09 ( ) 1.10 ( ) Hemiplegia, paraplegia, paralysis, functional disability (CC 67-69, , ) 1.06 ( ) 1.09 ( ) 1.06 ( ) 1.07 ( ) Cardio-respiratory failure or shock (CC 79) 1.15 ( ) 1.15 ( ) 1.16 ( ) 1.16 ( ) Congestive heart failure (CC 80) 1.19 ( ) 1.18 ( ) 1.20 ( ) 1.19 ( ) Acute coronary syndrome (CC 81-82) 1.12 ( ) 1.10 ( ) 1.10 ( ) 1.11 ( ) Coronary atherosclerosis or angina (CC 83-84) 1.05 ( ) 1.07 ( ) 1.06 ( ) 1.06 ( ) Valvular or rheumatic heart disease (CC 86) 1.08 ( ) 1.05 ( ) 1.07 ( ) 1.07 ( ) Specified arrhythmias (CC 92-93) 1.10 ( ) 1.12 ( ) 1.09 ( ) 1.10 ( ) Stroke (CC 95-96) 1.07 ( ) 1.05 ( ) 1.08 ( ) 1.06 ( ) Vascular or circulatory disease (CC ) 1.07 ( ) 1.07 ( ) 1.06 ( ) 1.06 ( ) Chronic obstructive pulmonary disease (CC 108) 1.17 ( ) 1.19 ( ) 1.17 ( ) 1.18 ( ) Fibrosis of lung or other chronic lung disorders (CC 109) 1.08 ( ) 1.09 ( ) 1.09 ( ) 1.09 ( ) Asthma (CC 110) 1.00 ( ) 0.99 ( ) 0.98 ( ) 0.99 ( ) Pneumonia (CC ) 1.06 ( ) 1.04 ( ) 1.07 ( ) 1.06 ( ) Pleural effusion/pneumothorax (CC 114) 1.10 ( ) 1.09 ( ) 1.12 ( ) 1.10 ( ) Readmission Measures Maintenance

39 Variable OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) Other lung disorders (CC 115) 1.03 ( ) 1.03 ( ) 1.04 ( ) 1.03 ( ) End stage renal disease or dialysis (CC ) 1.22 ( ) 1.16 ( ) 1.22 ( ) 1.20 ( ) Renal failure (CC 131) 1.18 ( ) 1.15 ( ) 1.16 ( ) 1.16 ( ) Urinary tract infection (CC 135) 1.06 ( ) 1.05 ( ) 1.08 ( ) 1.06 ( ) Other urinary tract disorders (CC 136) 1.03 ( ) 1.02 ( ) 1.04 ( ) 1.03 ( ) Decubitus ulcer or chronic skin ulcer (CC ) 1.14 ( ) 1.10 ( ) 1.09 ( ) 1.11 ( ) Vertebral fractures (CC 157) 1.13 ( ) 1.10 ( ) 1.07 ( ) 1.10 ( ) Other injuries (CC 162) 1.02 ( ) 1.06 ( ) 1.06 ( ) 1.05 ( ) Between Hospital Variance (SE) 0.026(0.002) 0.024(0.002) 0.027(0.002) 0.027(0.001) Readmission Measures Maintenance

40 Table 11 Pneumonia GLM Performance over Different Time Periods Characteristic c-statistic Predictive ability, % (lowest decile highest decile) Readmission Measures Maintenance

41 Table 12 Distribution of Hospital Pneumonia Volumes and RSRRs by Time Period Characteristic Number of Hospitals 4,826 4,813 4,761 4,925 Hospital Volume Mean (SD) 87.5 (81.4) 84.1 (77.1) 78.2 (73.3) (228.8) Range (min. max.) , th percentile th percentile th percentile RSRR (%) Mean (SD) 18.4 (1.1) 18.3 (1.0) 18.5 (1.1) 18.4 (1.6) Range (min. max.) th percentile th percentile th percentile Readmission Measures Maintenance

42 Figure 6 Distribution of 30 Day Pneumonia RSRRs in the Calendar Year Dataset N= 4,925 hospitals Readmission Measures Maintenance