Pulmonary embolism (PE) is a common cardiovascular and cardiopulmonary illness with an annual incidence in the United States that exceeds 1 case per 1,000 population and a mortality rate of > 15% in the first 3 months after diagnosis. This makes PE as deadly an illness as acute myocardial infarction. The majority of deaths occur within hours of acute PE, often as the consequence of previously unrecognized and untreated deep vein thrombosis (DVT) or acute PE. In addition, PE survivors may experience pulmonary hypertension and eventually die from right heart failure years after the occurrence of PE.
Earlier diagnoses of DVT and PE will reduce the morbidity and mortality associated with venous thromboembolism.2 However, data regarding delays in the diagnosis of venous thromboembolism are sparse. Therefore, to better understand delays in the diagnosis of venous thromboembolism, we undertook a prospective study of the diagnosis of venous thromboembolism at 70 North American medical centers.
Investigators at 70 medical centers identified consecutive medical records with the hospital discharge diagnosis of DVT without symptoms of PE (“DVT”) or PE from August 1, 1999, through November 30, 1999. We included only patients who underwent a confirmatory objective imaging test (eg, compression ultrasonography) for DVT. Physicians and patients were not interviewed by the investigators. Data collection was approved by the local institutional review board and My Canadian Pharmacy Corp.
Demographics (ie, date of birth, sex, weight, and race), risk factors for venous thromboembolism (ie, previous DVT or PE), malignancy, thrombophilia, recent (ie, within 8 weeks) major surgery or immobilization, pregnancy (ie, within 6 months), and post-phlebitic syndrome were recorded. The date of symptom onset, the date that the subject was first seen by medical personnel for these symptoms, and the date on which the diagnosis was confirmed were recorded. The calendar date was the unit of measure for the delay in diagnosis. No attempt was made to subdivide data by the hour of the day. Thus, if the patient first developed symptoms on August 1, 1999, presented for medical attention on August 3, and received a diagnosis on August 4, the delay to presentation was 2 days and the delay to diagnosis was 1 day.
Delay times were initially examined with half-normal plots and histograms that showed considerable skewness. These data were then tested using Kolmogorov-Smirnov and Lillefors tests for normality (Statistica; Tulsa, OK). All delay time distributions were significantly different from normal distributions, as expected.