Identifying relationships between process measures designated for the CMS Value-Based Purchasing program

CMS plans to link a portion of hospital payments to the quality of care provided to Medicare beneficiaries through a Value-based Purchasing (VBP) initiative. In its proposed roll-out, VBP will evaluate hospitals using a composite of 17 of the National Hospital Quality Measures. These measures include seven acute myocardial infarction (AMI), three heart failure (HF), five pneumonia (PN), and two surgical care improvement project (SCIP) measures. Traditionally, performance evaluation has focused on one clinical area at a time using composite scores. To our knowledge, there have been no studies examining the correlations between measures beyond their clinical groupings. Yet, understanding and leveraging such correlations may prove useful in designing interventions that help improve performance across measure sets. We used performance data from 331 hospitals participating in the Quality Indicator (QI) Project® during 2007 . Data for two of the 17 measures (Thrombolytic agent received within 30 minutes of hospital arrival and Primary percutaneous coronary intervention (PCI) received within 90 minutes of hospital arrival) were removed from analysis as not many hospitals participated in these measures. A variable clustering method was used to find groups of measures that are as correlated as possible among themselves and as uncorrelated as possible with measures in other clusters in terms of performance. Cluster means were compared using ANOVA methodology. The analysis revealed five clusters. (1) measures for medication other than antibiotics on admission and discharge, (2) medication other than antibiotics administered during in-hospital hospitalization, (3) antibiotics, (4) counseling/discharge planning, and (5) vaccination measures clustered together. Hospital's performance in measures relating to medications given at arrival and discharge were on the average, 11 percents higher than with vaccination related measures (P<0.05). There was no statistical difference in hospital's performance in measures relating to counseling/discharge planning, antibiotics, and medications administered during in-hospital hospitalization (P > 0.05). This study provides a context for understanding clinical measures by revealing the five dimensions of care. The results suggest that grouping measures by type of intervention may be just as meaningful as by clinical area.