Commensurate ratings of health care

A new descriptive item statistic, termed the mean cumulative logit (MCL), is advocated for scoring ratings of health care at the population level. The advantages of the MCL are demonstrated on data from the Consumer Assessment of Health Plans Study (CAHPS). The CAHPS data require (1) the comparison of binary and ordinal ratings in a common metric, (2) a treatment of unit and item nonresponse, and (3) the control of ordinal item correlations. These requirements are handled by a cumulative logit model that is applicable to unweighted (incomplete) and weighted (complete) data. The former case gives item satisfactions from a patient perspective. The latter case generates these satisfactions as social utilities from a provider perspective. From both of these viewpoints the perceived quality of health care is greater for fee-for-service plans than managed care plans in the field test population studied here.