POEMs and Tips from Other Journals

Previous | Next

Clinical Prediction Rule for Stratifying Patients with PE

More than 100,000 persons in the United States were hospitalized with pulmonary embolism (PE) in 2002. These hospitalizations resulted in 676,700 inpatient days. Among patients with PE, the short-term mortality rate varies from less than 2 percent for those with nonmassive PEs to 95 percent for patients with cardiorespiratory arrest. Models for the risk stratification of PE are not well established, and those in place are dependent on arterial blood gas analysis and leg vein ultrasound. Aujesky and colleagues set out to develop a purely clinical prediction rule for stratifying the risk of mortality or other adverse medical outcomes in patients with PE with the intent that low-risk patients can be treated in an outpatient setting.

A study population of 15,531 was developed from patients added to the Pennsylvania Health Care Cost Containment Council database from January 2000 to November 2002. The population was based on a primary discharge diagnosis of PE or a secondary diagnosis of PE combined with a primary diagnosis of cardiorespiratory failure, cardiogenic shock, cardiac arrest, secondary pulmonary hypertension, syncope, thrombolysis, or intubation/mechanical ventilation. The prediction rule was derived from 11 clinical variables routinely available at presentation. These variables included demographic data, comorbidities, and clinical and laboratory findings. Using logistic regression with 30-day mortality as the outcome, the authors found 11 clinical variables that were independently associated with 30-day mortality (see accompanying table). These results were validated internally and externally. The prediction rule was internally validated using 5,177 (33 percent) of the original study population. The external validation used data from 221 patients diagnosed with PE via spiral computed tomography at three university hospitals in Switzerland and one in France. Because the results from both validations were consistent with the results of the original study, the 11-variable prediction rule was highly reliable.

Each of the 11 factors was given a score to quantify the magnitude of mortality (see accompanying table). When the laboratory findings were incorporated into the model with the 11 clinical findings, all the findings except heart failure were still independently associated with 30-day mortality. Mortality increased stepwise with increasing risk class in all examples. The study's rule accurately identified patients who were at low risk for fatal and nonfatal outcomes. The ability to identify low-risk patients further supports recent evidence that many patients with nonmassive PE can be safely treated as outpatients using low-molecular-weight heparin.

The authors conclude that the use of a practical clinical risk assessment tool that accurately classifies patients with PE from low risk to high risk has the potential to decrease hospital stays and lower patient care costs. The authors also state that this tool needs further testing before it can be incorporated into practice.

JILL GIORDANO, MSIV

University of Medicine and Dentistry of New Jersey School of Osteopathic Medicine

Aujesky D, et al. Derivation and validation of a prognostic model for pulmonary embolism. Am J Respir Crit Care Med October 15, 2005;172:1041-6.