The CAHPS Health Plan Survey has probably been tested more completely than any previously used consumer survey.
There are two different and complementary approaches to assessing the reliability and validity of a questionnaire (1) cognitive testing, which bases its assessments on feedback from interviews with people who are asked to react to the survey questions, and (2) psychometric testing, which bases its assessments on the analysis of data collected by using the questionnaire. Although many existing consumer questionnaires about health care have been tested primarily or exclusively using a psychometric approach, the CAHPS team views the combination of cognitive and psychometric approaches as essential to producing the best possible survey instruments. Consequently, both methods have been included in the development of the CAHPS survey.
The cognitive testing method provided useful information on respondents' perceptions of the response task, how respondents recalled and reported events, and how they interpreted specified reference periods. It also helped identify words that could be used to describe health care providers accurately and consistently across a range of consumers (e.g., commercially insured, Medicaid, fee-for-service, managed care, lower socioeconomic status [SES], middle SES, low literacy, higher literacy) and helped explore whether key words and concepts included in the core questions worked equally well in both English and Spanish.
The CAHPS consortium also tested each CAHPS reporting composite in focus groups with plan members. Cognitive interviews with consumers were conducted to ensure that the reporting composites and their labels were easily understood. Psychometric analyses using data collected during pilot tests were also conducted. These analyses indicated that both the composites and the items in each composite were reliable and valid measures of members' experiences. In addition, items in each reporting composite were tested and found to be internally consistent. For example, reliability coefficients (Cronbach's alpha) in one pilot test involving four health plans using the instrument that most resembled the final CAHPS 2.0 instrument ranged from a low of 0.68 for the "Getting Needed Care" composite to a high of 0.90 for the "How Well Doctors Communicate" composite. These composites are positively associated with members' ratings of overall care provided by doctors and nurses and ratings of health plans.
In addition, the CAHPS development team, together with researchers from the National Committee on Quality Assurance (NCQA), conducted a detailed comparative analysis of the items in the CAHPS questionnaire and NCQA's Member Satisfaction Survey (MSS) from the fall of 1997 to the spring of 1998. These questionnaires were merged to form the 2.0 version of the CAHPS questionnaire. This testing is noteworthy because it was so extensive and because of the wide array of techniques used. These included focus groups, in-depth cognitive testing, pilot studies, methodological experiments, and large demonstration studies, such as the demonstrations in Washington State, Kansas, and New Jersey. NCQA also worked with the CAHPS consortium to conduct field tests of the 4.0 instrument with six health plans in spring 2005.