Construct validity comparisons of three methods for measuring patient compliance.

A multitrait - multimethod design was employed to assess the construct validity of three commonly used methods for assessing patient compliance: physiological assessments (e.g., blood chemistries), ratings by health professionals, and patient self-reports. Subjects were patients receiving ambulatory hemodialysis treatments for end-stage renal disease, whose regimen required them to take medications, to follow dietary restrictions, and to limit fluid intake. Study findings indicated that of the three methods examined, the nurse rating approach was the most valid (although it contained only about 50 percent valid variance). Measures derived from physiological assessments contained a substantial proportion of residual error (over 70 percent), and the patient self-report method contained only about 12 percent valid variance (with about 18 percent method-effects variance, and 68 percent residual-error variance). These results make clear the need for additional research directed at developing valid methods for evaluating patient compliance behaviors.