NEW YORK (Reuters Health) – Clinical prediction rules for judging the likelihood of a pulmonary embolism (PE) could be improved, new research shows.

Now the rules leave out certain variables that could help raise or lower suspicion, according to the authors. In their analysis, two of the strongest predictors of PE were a history of non-cancer-related thrombophilia and pleuritic chest pain.

The goal of the study was to assess the predictive value of information that physicians commonly collect from patients with suspected PE, information that has not been formally validated, Dr. Jeffrey A. Kline from Carolinas Medical Center, Charlotte, North Carolina told Reuters Health by email.

Dr. Kline and his colleagues examined 13 “implicit variables” that are commonly used to initiate, delay, or obviate testing for PE, yet have not been validated or incorporated into existing prediction rules. The implicit variables were compared with 12 “explicit variables” that are included in prediction rules.

In a paper published online January 4th in the Annals of Emergency Medicine, the researchers report that they used data from 7940 emergency room patients who had formal testing for pulmonary embolism, as ordered by 477 different clinicians. Eventually, 568 patients (7.2%) met standard criteria for pulmonary embolism or deep vein thrombosis.

Among the implicit variables, a history of non-cancer-related thrombophilia (OR, 1.99), pleuritic chest pain (OR, 1.53), and family history of venous thromboembolism (OR, 1.51) were positively associated with venous thromboembolism, while female gender (OR, 0.57), current smoking (OR, 0.59), and substernal chest pain (OR, 0.58) were negative predictors. The presence of tachypnea and patient perception of dyspnea were marginally associated with an increased risk of venous thromboembolism.

Four of the implicit variables commonly used to support formal PE testing proved to be statistically nonsignificant, including pregnancy or postpartum state, sudden onset of symptoms, obesity, and history of treated but currently inactive malignancy.

Among the 12 explicit variables, 3 (hemoptysis, trauma within 4 weeks, and shock index greater than 1.0) were not statistically associated with venous thromboembolism. Of the remaining 9 variables, the strongest predictors were patient history of venous thromboembolism (OR, 2.90) and unilateral leg swelling (OR, 2.60), surgery in the last 4 weeks (OR, 2.27), current estrogen use (OR, 2.31), and hypoxemia (OR. 2.10).

Dr. Kline said that the current findings could have legal implications.

“The variables we studied are very often the subject of debate between opposing expert witnesses in medical malpractice lawsuits alleging negligence in failure to diagnose PE,” he explained. “I hope this publication can objectivize some of these arguments.”

Dr. Kline added that he plans to investigate the use of a computerized model to produce a numeric pretest probability based on these findings.

Ann Emerg Med 2009.