Analyzing Patient Case Mix and Hospital Rankings

To the Editor: In their study of patient case-mix adjustment, hospital process performance rankings, and eligibility for financial incentives, Dr Mehta and colleagues¹ tackled an important question in health services research. However, the unit of analysis they used in their study seems inconsistent. For example, in the Methods section, the authors stated that the analysis used opportunity-based data. Later, however, the authors stated that they adjusted for opportunity mix in the analyses. It seems that adjustment should be performed at the patient level, not the opportunity level. Furthermore, the authors stated that the hospital rankings were based on the scores for composite process measures. This again indicates the unit of analysis should be patient, not opportunity (ie, quality measure), because each patient has a composite score summarizing 8 quality measures.²

Traditionally, process measures are often modeled at the hospital level.³ In the study by Mehta et al, to adjust for patient mix and rank hospitals, the authors seemed to adopt the approach for modeling outcome measures, in which analyses are often performed at the patient level.⁴ However, they modified the outcome-oriented approach by using opportunity as the unit of analysis; ie, a patient would have multiple observations in the data set if the patient was eligible for multiple measures. If this is the case, an additional level may be needed in the hierarchical model to account for clustering of opportunities within patients.

The choice of analytic approach might influence the results of hospital rankings. Thus, the issue of unit of analysis used in this article requires clarification.