Alkaline phosphatase isoenzyme determination (APID) is in common use despite evidence suggesting that the results correlate poorly with actual sites of disease. To assess the predictive value of this test in clinical practice, 99 APIDs performed on 94 patients were identified and the patients' charts were reviewed. Results of APID were compared with actual patient diagnoses as determined by other means. The liver isoenzyme fraction was not very accurate in predicting the presence of liver disease (positive predictive value 68%). In contrast, the bone isoenzyme fraction was insensitive (56%) but a positive test predicted bone disease well (positive predictive value 93%). The association of elevated transaminases with elevated alkaline phosphatase on a chemistry profile was as useful as APID in identifying liver disease, suggesting that APID should not be done in this setting. Using this information, APID can be helpful in the assessment of an ill patient with an elevated alkaline phosphatase.