How valid and reliable are patient satisfaction data? An analysis of 195 studies

Objective. To assess the properties of validity and reliability of instruments used to assess satisfaction in a broad sample of health service user satisfaction studies, and to assess the level of awareness of these issues among study authors. Design. Examination and analysis of 195 papers published in 1994 in 139 journals. The following databases were searched: British Nursing Index, CINAHL, EMBASE, Medline, Popline, and PsycLIT. Main measures. Number and types of strategies used for content, criterion, and construct validity, and for stability and internal consistency. Associations between validity/reliability and other study characteristics. Results. Eighty-nine (46%) of the 195 studies reported some validity or reliability data; 76 reported some element of content validity; 14 reported criterion validity, with patient's intent to return the most commonly used criterion; four reported construct validity. Thirty-four studies reported internal consistency reliability, 31 of which used Cronbach's coefficient a; eight studies reported test-retest reliability Only 11 studies (6% of the 181 quantitative studies) reported content validity and criterion or construct validity and reliability. 'New' instruments designed specifically for the reported study demonstrated significantly less evidence for reliability/validity than did 'old' instruments. Conclusion. With few exceptions, the study instruments in this sample demonstrated little evidence of reliability or validity Moreover, study authors exhibited a poor understanding of the importance of these properties in the assessment of satisfaction. Researchers must be aware that this is poor research practice, and that lack of a reliable and valid assessment instrument casts doubt on the credibility of satisfaction findings.