The methylene blue stain for fecal leukocytes (FL) is widely used as an adjunct to slower but more accurate tests of diarrheal etiology, such as stool culture (SCx) or toxin assays for Clostridium difficile. Prior studies investigating the utility of FL for predicting SCx and C. difficile toxin assay (CDTA) results did not evaluate the importance of inpatient versus outpatient status. We conducted a study of patients who submitted a stool specimen to the Stanford Hospital Microbiology Laboratory between May 1998 and April 1999. The results for stool specimens that were tested by FL and by a confirmatory test (either SCx or CDTA) were used to determine whether the FL method helped to predict the results of these tests. Of 797 stools that were tested by FL method and at least one confirmatory test, 502 stools were tested by CDTA, and 473 stools were cultured. The FL test was 14% sensitive and 90% specific for C. difficile with a diagnostic threshold of one white blood cell/high-power field (WBC/HPF). The overall likelihood ratio (LR) for a positive CDTA was 1.4 with a 95% confidence interval (CI) of 0.5 to 3.7 (P = 0.5) and was similar among inpatients and outpatients. In contrast, the presence of 1 WBC/HPF was 52% sensitive and 88% specific for the 27 positive SCx results and helped to predict a positive SCx result (LR, 4.2; 95% CI, 2.7 to 6.5; P < 0.001). The sensitivity of 1 WBC/HPF was 57%, and its predictive value for SCx was higher among outpatients (outpatient LR, 5.0; 95% CI, 2.9 to 8.6; P < 0.001; inpatient LR, 1.9; 95% CI, 0.3 to 10.8; P = 0.5). Among inpatients, only 4 (1.5%) of the 273 SCx results were positive, and the presence of 1 WBC/HPF was insensitive (25%) and did not predict a positive SCx (LR, 1.9; 95% CI, 0.3 to 10.8; P = 0.5). When the data were reanalyzed using a diagnostic threshold of five WBC/HPF for FL, the predictive power of the FL method was similar. Thus, FL was of no value in predicting CDTA positivity, nor was it helpful in predicting SCx results for inpatients. Neither SCx nor the FL method should routinely be performed on samples from inpatients. Among outpatients, presence of FLs should suggest a bacterial diarrhea in clinically compatible cases.