Effects of method error on the power of a statistical test. Implications of imperfect sensitivity and specificity in...

Errors in classifying individuals as to whether they have a certain characteristic may adversely affect the power of a statistical test to detect differences in the incidence of that characteristic between groups or to detect changes in the incidence over time. To determine whether the classification method to be used in the Study on the Efficacy of Nosocomial Infection Control (SENIC Project)--retrospective chart review--would provide sufficient power for a test of the study's main hypothesis, the authors calculated power as a function of the sensitivity and specificity of the method. They then contrasted the case in which sensitivity and specificity are constant for all hospitals in the study with cases in which the measures vary between groups of hospitals or between time periods. It was found that variation within the observed ranges of sensitivity and specificity, even if systematic, has little effect on power unless it is related to the hypothesis under study. The authors thus concluded that SENICs study design is adequate to detect any substantial success of infection surveilliance and control programs in combatting nosocomial infection, unless such programs affect the accuracy of the chart review method.