Strategies for Researching the Impact of Pay for Performance Programs II

CMS Premier Hospital Quality Incentive Demonstration (PHQID) Evaluation Linda A. Radey Linda.Radey@cms.hhs.gov Office of Research, Development, and Information June 4, 2007 • 3-year hospital P4P demonstration October 1, 2003 through September 30, 2006 (extended 3 additional years) • 267 hospitals participated – Less than 10 percent disenrolled • 34 quality measures – all-patient data – 27 process measures – 7 risk-adjusted outcome measures 5 Clinical Areas • Acute Myocardial infarction (AMI) • Heart failure (HF) • Coronary artery bypass graft (CABG) • Pneumonia (PN) • Hip and knee replacement (HK) Compute Composite Quality Scores (CQS)--Roll up measure scores in each area. Rank hospitals by CQS in each clinical area. Incentives: •Hospitals with CQS in top decile— 2 percent bonus •Hospitals with CQS in 2nd decile— 1 percent bonus Penalties at end of year 3 for hospitals with CQS below thresholds established at end of year 1: •1 percent for CQS below 9th decile threshold •2 percent for CQS below 10th decile threshold Evaluation Research Questions – Did the PHQID increase the quality of hospital care beyond any increases that were the result of P4R & other quality improvement programs operating during the demonstration? (Quality Analysis) – Did the PHQID affect Medicare reimbursements to acute care hospitals enrolled in the demonstration? (Medicare Reimbursement Analysis) Possible Comparison Groups • 3,500 non-participating acute care hospitals (including nonparticipating PerspectiveTM hospitals) • 146 non-participating PerspectiveTM hospitals • Compare all PerspectiveTM participating & non-participating hospitals with all other hospitals Quality Analysis • Limited availability of data could cause demonstration effect to be underestimated. Baseline quality data not used. Limited to 3 clinical areas (AMI, HF, & PN) for some participating hospitals. – Comparison hospital quality data available for subset of PHQID measures & data unavailable for 1st quarter of the demonstration. (Hospital Compare) Quality Analysis (contd.) • Demonstration quality measures were not constant—difficult or impossible to determine quality improvement. – 3 measures were revised to reflect current clinical evidence & keep them aligned with CMS, Joint Commission, NQF and STS counterpart measure definitions. – 4 measures were suppressed for some portion or entire demonstration. Quality Analysis (contd.) • Processing quality data for participating hospitals takes an average of 15 months from the end of the demonstration year— creates delays for quality & reimbursement analysis. • Hospital level quality data used for participating & comparison hospitals--- patient-level quality data not available. Medicare Reimbursement Analysis • Measure effects on Medicare acute care hospital reimbursement. • Data source: patient-level stay data from MedPAR. • Baseline data available-use 2 years prior to demonstration. • Analysis design based on an episode of care: – Starts when patient admitted for any of the 5 clinical areas included in the demonstration. – Includes any subsequent admissions for any condition to any acute care hospital that occurs within 30, 60, or 90 days after discharge for initial stay. Medicare Reimbursement Analysis (continued) Medicare Reimbursement Analysis (continued) • DRG weight =unit of measurement for reimbursement. – Excludes: regional payment differences, DSH payments, IME, outlier payments, and passthrough payments. • Episode weight=DRG weight for initial stay + DRG weights for all subsequent stays. Medicare Reimbursement Analysis (continued) • Total episode $s = Total episode weight x national standardized payment rate. • Analysis will capture effects of higher quality on: fewer complications & decreased admissions. • Quality analysis---quality data not available for baseline period. – Rely on including correct covariates without check of model specification. – Construct quality measures from CareScienceTM indicators using administrative data for baseline & demonstration periods. Primary Difference between Quality & Reimbursement Analysis • Reimbursement analysis—data available for baseline & demonstration periods. – Allows check of model specification.