- How were scores adjusted for respondent characteristics?
We know that respondents with certain characteristics tend to give more favorable reports than respondents with other characteristics. For example, older respondents tend to give more favorable reports than younger respondents. So that individual physicians will not have their scores inflated, or deflated, simply because of the characteristics of the patients they serve, we calculated an "adjusted score" for each physician for each question, doing a case-mix adjustment that takes into account self-reported health status, age, and education level. These are the three characteristics approved for case-mix adjustment by NQF.
The case-mix adjustment is done as follows for each question for each doctor. A regression analysis is used to predict, for each respondent, the score someone with that respondent's characteristics would be expected, on average, to give a doctor. The mean of these predicted scores for each of a doctor's respondents is calculated to get a predicted score for that doctor for that question. Then the mean of the actual scores given by each respondent for that doctor for that question is calculated. Then the differencethe actual minus the predicted scorefor that doctor for that question is calculated. Finally, that difference is added to the mean score for all survey respondents for that question to get the adjusted score for that doctor for that question.