Adjusting for patient characteristics when analyzing reports from patients about hospital care

To determine which patient characteristics are associated with reports and ratings of hospital care, and to evaluate how adjusting reports and ratings for hospital differences in such variables affects comparisons among hospitals.

A telephone survey of a sample of patients hospitalized in 22 hospitals in a single city and a statewide mail survey of hospitalized patients.

The surveys assessed: respect for patients' preferences, coordination of care, information exchange between patient and providers, physical care, emotional support, involvement of family and friends, and transition and continuity. The surveys also asked patients to rate their doctors, nurses, and other hospital staff.

The variables with the strongest and most consistent associations with patient-reported problems were age and reported health status. Patient gender and education level also sometimes predicted reports and/or ratings. Models including these variables explained only between 3% and 8% of the variation in reports and ratings.

The impact of adjusting for patient characteristics on hospital rankings was small, although a larger impact would be expected when comparing hospitals with more variability in types of patients. Nevertheless, we recommend adjusting at least for the most important predictors, such as age and health status. Such adjustment helps alleviate concerns about bias. It also may be useful to present data for certain groups of patients (ie, medical, surgical, obstetric) separately to facilitate interpretation and quality improvement efforts.