OUTCOME OF AIDS PATIENTS REQUIRING MECHANICAL VENTILATION PREDICTED BY RECURSIVE PARTITIONING

Citation
Aa. Elsolh et al., OUTCOME OF AIDS PATIENTS REQUIRING MECHANICAL VENTILATION PREDICTED BY RECURSIVE PARTITIONING, Chest, 109(6), 1996, pp. 1584-1590
Citations number
27
Categorie Soggetti
Respiratory System
Journal title
ChestACNP
ISSN journal
00123692
Volume
109
Issue
6
Year of publication
1996
Pages
1584 - 1590
Database
ISI
SICI code
0012-3692(1996)109:6<1584:OOAPRM>2.0.ZU;2-4
Abstract
Mechanical ventilatory support (VS) is often required for patients wit h AIDS. Patients and/or their surrogates often ask tile likely outcome of this intervention. To answer this question, we have developed a cl assification tree using clinical data from 71 patients with AIDS ident ified from the discharge abstracts of two hospitals between January 19 90 and September 1994. These data were obtained at the time of hospita l admission prior to any treatment and before VS was initiated. Surviv al was defined as discharge from the hospital that occurred in 13 of 7 2 admissions reviewed. A classification tree was developed by binary r ecursive partitioning. The output of the resulting tree was adjusted t o produce a positive predictive value for death of 100% (95% confidenc e interval [95% CI], 94 to 100%) and a sensitivity and specificity of 98% (95% CI, 91 to 100%) and 100% (95% CI, 74 to 100%), respectively. The negative predictive value was 92% (95% CI, 64 to 100%). The tree p redicted that patients with lactate dehydrogenase (LDH) levels less th an 1,176 IU/L survived until hospital discharge, unless they had a pos itive blood culture, active tuberculosis prior to VS, a blood CD4 coun t less than le cells per cubic millimeter, or creatinine and hemoglobi n values that were either above 2.4 mg/dL or less than 8.5 mg/dL, resp ectively. The remainder of the patients with an LDH level above 1,176 IU/L in this study died before hospital discharge. The classification tree requires prospective validation before it can be used as a predic tive instrument. Nevertheless, this approach can be used to develop a concise summary of the local outcome experience of this circumstance i n a manner that could be conveyed to patients and/or their surrogates.