A computer based intervention to reduce unnecessary serologic testing

Laboratory testing is important in the evaluation of patients with possible systemic rheumatic disease, but uncritical use of any test may result in misleading information and unnecessary costs. We attempted to reduce the number of unnecessary antinuclear antibody, rheumatoid factor, and complement level tests ordered by house officers at a large teaching hospital, where inpatient orders are written through a computer based order entry system.

We conducted a prospective cohort study of an interactive test ordering program. The intervention consisted of displaying post-test probability estimates during the usual physician order entry session. These estimates were based on pretest probabilities entered by the ordering physician and sensitivities and specificities derived from a literature review. Another group of test orders did not prompt the intervention and were considered controls. The outcome of interest was the percentage of tests canceled in the intervention group versus the control group.

Eleven percent (11/99) of intervention orders were canceled, versus only one order among 236 controls (p = 0.001). However, there was no association between the physicians' pretest probability estimates and whether test orders were canceled (p = 0.59). Additionally, 43 of the 335 orders (13%) yielded positive tests, but only 4 patients (1%) were given new diagnoses of rheumatic disease.

The computer based intervention significantly reduced orders for antinuclear antibody and rheumatoid factor levels by 10%. Further reductions without clinical harm are probably possible, since the yield of testing for new rheumatic diseases was low.