To develop and validate a multivariate model for predicting relapses after treatment of decompensated chronic obstructive pulmonary disease in an emergency department.
A 5-year survey was conducted, including training and validation periods. Stepwise logistic regression was used to develop a multivariate predictive model using clinical data obtained at the time of each visit. A relapse was defined as an unscheduled return to the emergency department within 48 hours.
The study was conducted in the emergency department of the Albuquerque (New Mexico) Veterans Affairs Medical Center.
The subjects were 289 patients with documented chronic obstructive pulmonary disease.
Measurements and Main Results.—
During the first 3 years, there were 705 visits in which the patient was treated and released from the emergency department. Relapse occurred 82 times (11.6%). Logistic regression showed that the following variables had an effect on the risk of relapse: the relapse rate for previous visits, a previous visit within 7 days, long-term home oxygen therapy, the number of doses of nebulized bronchodilators, the administration of aminophylline, and the use of antibiotics and prednisone at the time of discharge from the emergency department. During the next 2 years, the 48-hour relapse rate was 9.9% (47 of 476 discharges). When the model was fitted to these data, all of the original variables contributed to the prediction of relapse except antibiotic use and long-term home oxygen therapy. The logistic model was used to categorize each visit during the validation phase. The relapse rate for "high-risk" visits was significantly higher than that for "low-risk" visits (18.4% vs 6.1%). The method identified 57.4% of visits that ended in relapse at 48 hours.
A multivariate model can be used to identify patients with a poor prognosis after the outpatient treatment of decompensated chronic obstructive pulmonary disease.(Arch Intern Med. 1992;152:73-77)