In this study, the investigators determined whether the concentration of bilirubin obtained from aspirates of newly inserted feeding tubes improved the ability of pH alone to predict feeding tube location. An algorithm to predict feeding tube location was developed. Aspirates of gastrointestinal fluid from adult, acutely ill patients who had nasogastric (n = 209) or nasointestinal (n = 228) small-bore feeding tubes (8-F or 10-F) inserted were obtained within 5 minutes of radiographs confirming feeding tube placement. Patients were excluded if antacids or tube feedings had been administered within 4 hours preceding sample collection, any medications were administered orally or via feeding tube within 1 hour preceding sample collection, samples were grossly bloody, or if patients had a history of gastric surgery or trauma. Tracheobronchial secretions (n = 125) were obtained by suctioning patients who had an artificial airway; 24 pleural samples were obtained at the time of thoracenteses, and 1 sample was obtained from a feeding tube inadvertently positioned into the right mainstem bronchus. Approximately two-thirds of the patients were receiving H2 -receptor antagonists. The pH and bilirubin concentration of samples were measured using standard methods and compared with the radiographs of feeding tube location. Mean pH levels in the intestine (7.35 ± 0.6) and lung (7.73 ± 0.4) were significantly higher than in the stomach (3.9 ± .15). Mean bilirubin levels in the stomach (1.28 ± 0.25 and lung (.08 ± .02) were significantly lower than in the intestine (12.73 ± .91 mg/dL). To develop the predictive algorithm, the pH and bilirubin values were dichotomized by visually inspecting the overlap of the data distributions and mean differences by tube sites. A pH of 5 or less and bilirubin less than 5 mg/dL successfully identified more than two-thirds of the 209 feeding tubes radiographically confirmed as being in the stomach. Misclassification of 53 of the gastric feeding tubes as respiratory cases by this category was attributed to the effects of acid-inhibiting medications. Thirteen of the gastric tubes were misclassified as being in the intestine, most likely due to acid-inhibiting agents and reflux of intestinal fluid into the stomach. A pH greater than 5 and bilirubin of 5 or more correctly identified three-fourths of the 228 tubes in the intestine. Fifty-four of the intestinal tubes were misclassified as respiratory cases due to lower than expected bilirubin levels. A pH of less than 5 and bilirubin less than 5 mg/dL correctly classified 100% of the 150 respiratory cases. Approximately one-fourth of the 437 gastrointestinal tube placements were misclassified as respiratory cases by this category. Use of the algorithm was estimated to reduce the number of x-rays to exclude respiratory placements in 28% of the cases.