Diagnostic tests often fail to lead to changes in patient outcomes

Abstract

Objectives: To evaluate the effects of diagnostic testing on patient outcomes in a large sample of diagnostic randomized controlled trials (D-RCTs) and to examine whether the effects for patient outcomes correlate with the effects on management and with diagnostic accuracy.

Study design and setting: We considered D-RCTs that evaluated diagnostic interventions for any condition and reported effectiveness data on one or more patient outcomes. We calculated odds ratios for patient outcomes and outcomes pertaining to the use of further diagnostic and therapeutic interventions and the diagnostic odds ratio (DOR) for the accuracy of experimental tests.

Results: One hundred forty trials (153 comparisons) were eligible. Patient outcomes were significantly improved in 28 comparisons (18%). There was no concordance in significance and direction of effects between the patient outcome and outcomes for use of further diagnostic or therapeutic interventions (weighted κ 0.02 and 0.09, respectively). The effect size for the patient outcome did not correlate with the effect sizes for use of further diagnostic (r = 0.05; P = 0.78) or therapeutic interventions (r = 0.18; P = 0.08) or the experimental intervention DOR in the same trial (r = -0.24; P = 0.51).

Conclusion: Few tests have well-documented benefits on patient outcomes. Diagnostic performance or the effects on management decisions are not necessarily indicative of patient benefits.

Keywords: Diagnostic accuracy; Diagnostic odds ratio; Diagnostic tests; Empirical assessment; Patient outcomes; Randomized controlled trials.