UBC statistician picks apart popular algorithm used to screen chlamydia

The US Centre for Disease Control (CDC) has recognized UBC statistician Liangliang Wang for her research questioning the accuracy of a popular algorithm used in chlamydia screening tests.

Wang and her co-authors used computer simulated scenarios to test the accuracy of the patient estimation algorithm (PISA). The PISA is used to produce sensitivity and specificity estimates for Chlamydia trachomatis screening results, and is widely published and included in Food and Drug Administration-approved materials.

Wang’s team discovered that PISA can produce inaccurate estimates, because it assumes the measures it utilizes are perfect. However, tests designed to detect the presence of a bacterium can mistakenly detect a previous infection. False positives may also occur due to contamination. With such flawed data, the PISA produces inaccurate estimates which are in turn employed as the basis for screening tests. The CDC awarded the Charles C Shepard Science Award to Wang and her team for the discovery.

“There are more sophisticated measurement tools in statistics that could be employed,” says Wang. “The PISA is very simple but not very accurate.”

Wang and her co-authors hope their study will prompt a re-evaluation of the use of PISA. A multiple latent variable model — widely used in statistics — would yield more accurate results, says Wang.