| Summary: | This study examined 1265 consecutive patients and uniquely based the scan diagnoses on prior operational definitions of different scan patterns: it also uniquely determined the reproducibility (reliability) of the operational definitions. Diagnoses based solely on scan patterns were reliable. The accuracy of scanning was comparable to other published series. However, with the exception of the pattern for multiple metastases and some of the rarer more complex patterns, the diagnostic accuracy was probably too low for clinical practice. The diagnostic accuracy of scanning improved when the scan was considered in conjunction with simple clinical parameters such as of pain of an history of malignancy; this increases the prevalence of disease in the study population and improves diagnostic accuracy. The variables age, sex, ingestion of steroids, and the ESR and alkaline phosphatase were of little value. When scan diagnoses were considered in conjunction with radiology, there was a small but significant improvement in diagnostic accuracy. The data were then analyzed using discriminant analysis; this technique eliminated the problems of cross correlation. The nature of pain was found to be useful discriminant. In malignancy and infection, scanning was the most discriminating variable. In arthritis, trauma and Paget's disease, radiology was the most discriminating variable. The final task was to perform a cost-effective analysis. The most cost- effective use of resources occurred when scanning and radiology were utilized in conjunction with discordance between of symptoms and the original imaging modality. Having identified potential savings, it was difficult to realize them because fixed costs comprised a large proportion of the total costs. It was then estimated that as the costs of investigation significantly outweigh the costs of standard surgical treatment efficient investigation would release more resources than the avoidance of inappropriate surgery.
|
|---|
