Diagnostic tests are developed to be quicker, cheaper or less invasive than a reference test (or 'gold standard'), against which they are evaluated. The evaluation process is often known as validation of a diagnostic test. Long established in epidemiology, these techniques have found wider applications in other areas such as medical screening, pattern recognition systems (e.g. car model identification), drug impairment (of car drivers) and software reliability.
This course is a combination of presentations and computer-based practicals, whereby theory is firmly placed into practice. A variety of examples will be used and you will have a choice of statistical packages from SAS, SPSS and Stata.
Who should attend?
Scientists and technologists who wish to be conversant with the process of evaluation of diagnostic tests.
Familiarity with standard errors and confidence intervals for proportions is required. Knowledge equivalent to topics in our course A Review of Basic Statistics would be ideal.
How you will benefit
Participants will learn about standard measures and techniques that are used to evaluate the performance of diagnostic tests that yield a yes/no outcome. You wil also learn how to extend these techniques to diagnostic tests that yield categorical or continuous outcomes.
- Measures of performance for diagnostic tests yielding a yes/no outcome: sensitivity, specificity, predictive values, odds ratios and percent agreement
- Measures of performance for diagnostic tests yielding a categorical or continuous outcome: receiver operating characteristic (ROC) curves and area under the curve (AUC)
- A brief consideration of current international standards
The cost of the course is £305