Problems in the Use of Large Data Sets to Assess Effectiveness

Abstract

Large data sets are an attractive source of information for outcome assessment, but their use involves certain problems and risks. Data base evaluations are retrospective and unblinded; they often represent the result of multiple analyses of multiple endpoints, and it is difficult to identify the procedures used and analytic choices made because critical details are often omitted. While data bases can suggest problems and offer answers, they cannot prove them; data base analyses must be followed by trials.