Screening for clinical laboratory errors with Medicare claims data: results for digoxin.

A statistical model is demonstrated for finding outpatient clinical laboratories having high frequencies of bad patient outcomes, after testing, that may indicate testing errors. A sample of Medicare Part A and Part B claims for 30,685 digoxin tests for 1985 through 1987 was analyzed. Deaths or digoxin-related hospitalizations within 14 days of digoxin testing are defined as "adverse events" potentially influenced by test information. Approximately 3.3% of digoxin tests were followed by adverse events with two-thirds resulting in hospitalization or death. Adverse events were (a) lower in states with stronger laboratory regulations, (b) 15% higher whenever the site of testing switched, (c) unrelated to testing in low volume physician office laboratories, and (d) unrelated to frequency of digoxin testing in previous 6 months. Results a and b are consistent with findings for prothrombin testing but c and d are not. Differences are consistent with technical characteristics of the tests.