L Ketai1, T Washington, T Allen, J Rael. 1. Department of Radiology, University of New Mexico, Health Sciences Center, Albuquerque 87131-5336, USA.
Abstract
RATIONALE AND OBJECTIVES: The purpose of this study was to use logistic regression to analyze both Gram stain results and other clinical information to create a decision rule capable of predicting which abdominal or pelvic fluid collections would later prove to be infected and therefore require catheter drainage. MATERIALS AND METHODS: The authors retrospectively collected Gram stain results and clinical data (postoperative status and antibiotic use) regarding 124 abdominal or pelvic fluid drainage procedures performed between 1991 and 1996. They then analyzed these data by using logistic regression to create an equation that predicted the presence of fluid infection. Finally, they validated this equation by applying it to 39 abdominal or pelvic fluid drainage procedures performed in 1997. RESULTS: The resulting equation predicted that a fluid collection was likely to be infected if any of the following were present: Gram stain positive for bacteria, Gram stain showing moderate or many white blood cells, and purulent fluid at visual inspection. For the initial data set, the sensitivity of the decision rule was 91%, the specificity was 54%, and the overall accuracy was 77%. For the 1997 data set, the sensitivity of the decision rule was 88%, the specificity was 50%, and the accuracy was 77%. CONCLUSION: When combined with clinical information, Gram stain results are sensitive but nonspecific in the detection of abdominal or pelvic fluid infection. Use of the decision rule could prevent unnecessary catheter placement in a minority of patients with abdominal or pelvic fluid collections.
RATIONALE AND OBJECTIVES: The purpose of this study was to use logistic regression to analyze both Gram stain results and other clinical information to create a decision rule capable of predicting which abdominal or pelvic fluid collections would later prove to be infected and therefore require catheter drainage. MATERIALS AND METHODS: The authors retrospectively collected Gram stain results and clinical data (postoperative status and antibiotic use) regarding 124 abdominal or pelvic fluid drainage procedures performed between 1991 and 1996. They then analyzed these data by using logistic regression to create an equation that predicted the presence of fluid infection. Finally, they validated this equation by applying it to 39 abdominal or pelvic fluid drainage procedures performed in 1997. RESULTS: The resulting equation predicted that a fluid collection was likely to be infected if any of the following were present: Gram stain positive for bacteria, Gram stain showing moderate or many white blood cells, and purulent fluid at visual inspection. For the initial data set, the sensitivity of the decision rule was 91%, the specificity was 54%, and the overall accuracy was 77%. For the 1997 data set, the sensitivity of the decision rule was 88%, the specificity was 50%, and the accuracy was 77%. CONCLUSION: When combined with clinical information, Gram stain results are sensitive but nonspecific in the detection of abdominal or pelvic fluid infection. Use of the decision rule could prevent unnecessary catheter placement in a minority of patients with abdominal or pelvic fluid collections.