Relations between Value-based Purchasing (VBP) scores components

Background: In January 2011 CMS released the proposed methodology for VBP putting 1% of baseline DRG payment at risk for fiscal year 2013. Hospitals will be scored on improvement and achievement on 17 clinical measures and 8 satisfaction measures. Satisfaction measures contribute 30% and clinical measures contribute 70% to the overall VBP score. Data: VBP scores calculated from the most recent publicly available data using 2Q09 to 1Q10 for the assessment period and the prior year for the baseline period. Methods: we performed a cluster analysis to find natural subgroups of measures closely related to each other but different from other groups. We used a partition based algorithm classifying measures into classes explaining maximum variation, and an agglomerative approach that defined clusters consisting of measures with maximum correlation. Results: As expected, the clinical and satisfaction measures defined two natural groups. The variation analysis indicated no subgroups for the satisfaction measures and classified the clinical measures largely along the focus sets of AMI, pneumonia, heart failure and surgery (SCIP). The correlation analysis showed that, for instance, within the satisfaction cluster, communication with nurses and responsiveness of clinical staff were closely related. In the clinical domain, the use of fibrinolytic therapeutic within 30 minutes of arrival stands out as not being related to any other clinical measure. Other findings: Discharge information scores are relatively uncorrelated to other satisfaction measures. Conclusion: These results can help design more focused and effective quality improvement interventions by informing how improvements in one measure may affect another.