Online Program

Variation in quality by hospital characteristics and the implications for risk-adjustment

Alex Bohl, Mathematica Policy Research 
David Jones, Mathematica Policy Research 
*Dmitriy Poznyak, Mathematica Policy Research 
Jessica Ross, Mathematica Policy Research 
Eric Schone, Mathematica Policy Research 
Frank Yoon, Mathematica Policy Research 
Joe Zickafoose, Mathematica Policy Research 

Keywords: Hospital quality; risk-adjustment; hospital characteristics; peer grouping; CART analysis; regression

The academic literature and popular press have called into question the validity of the AHRQ Quality Indicators in making comparisons of hospital quality. A key component of the critiques is the assertion that certain hospital types are fundamentally different in their mission, patient populations, and service delivery, raising the question whether risk adjusting the QIs using only patient-level factors is sufficient to support hospital comparisons. Although research has documented differences in QI rate across hospital groups, these analyses were not designed to assess whether the QIs are biased against certain hospital types. We use the classification and regression trees’ (CART) method followed by regression analysis to explore whether hospitals’ structural characteristics—such as volume, scope, ownership, urbanicity and geographical region—drive the variation in hospitals’ QI rates. We then model the distribution of the QI rates by grouping hospitals by similar attributes (“peer groups”). Finally, we discuss how incorporating hospitals characteristics into the risk-adjustment model—either by smoothing QI rates to a peer-group’s risk adjusted mean or by adding fixed effects for hospital characteristics to the risk-adjustment model—may reduce the bias when comparing hospitals on their estimated QIs.