Shared patient analysis: a method to assess the clinical benefits of patient referrals.

Background: Referral to specialized physicians or institutions often is deemed necessary in clinical medicine, but no method exists to assess the clinical benefit of such referrals.
Objectives: To describe a method, which is shared patient analysis, to measure the expected improvement in clinical management associated with referrals and to apply that method in the field of abdominal and pelvic oncological radiology.
Subjects: All patients referred, during a 4-year period, to surgical oncologists at four academic centers (the referral providers, or RPs) with radiographs performed before referral at a community site (the initial providers, or IPs). Patients (n = 396) for whom both the IP interpretation and a final diagnosis was available were eligible. All IP and RP readings were placed in random order and presented to surgical oncologists, who then recommended a treatment course.
Measurements: Diagnostic accuracy of the IP and RP readings and the proportion of patients who were assigned to an appropriate treatment by the oncologist were determined.
Results: When the indication for imaging was primary diagnosis or staging, the kappa for presence of cancer was 0.70. When the indication was cancer follow-up, the kappa for presence of recurrent/progressing cancer was 0.66. There were disagreements between the IP and RP radiologists over the interpretation of 162 films, with the RP radiologists being correct in 153 (94%). Had the patients been treated using IP readings, there would have been 19 more inappropriate surgeries and 19 more admissions (both P <0.05) than if the oncologists had based their recommendations on RP readings.
Conclusions: The technique of shared patient analysis permits assessment of the clinical benefits associated with referrals.