Surveillance of severity of infection
This site describes, and contains source code for, a tool for monitoring the severity of infection.
We were interested in monitoring changes in continuous variables over time, such a biomarkers obtained on diagnosis of infection.
It is known that some biomarkers (such as white cell count) reflect severity of illness. As such, if illness severity increases in a patient group, the biomarker may change, and may (we have shown) provide early warning of changing infection severity. For example, consider the situation where some biomarker remained constant for 2 years in a particular patient group, then rose for the next 4 years, then remained static (example). Early detection of the change would be desirable clinically. The technique developed attempts to find changepoints in such data, allowing early detection of change (example).
It is based on joint-point based regression method ('iterative sequential regression', or ISR) in which a range of models are automatically compared and the best fitting one selected, using information criteria. Both quantile and general linear models can be used.
An implementation in R is available.
The technique was developed with studies of C. difficile colitis.
Please contact the authors, below.
The technique was developed by Iryna Schlackow, Sarah Walker and David Wyllie with support from the National Institutes for Health Research.
The work was based in Oxford University Hospitals Biomedical Research Centre Infection Theme and used data from the Infection in Oxfordshire Research Database.
For contact details, please use our contact form.
Please cite this article Schlackow I et al (2012) Plos Medicine http://dx.doi.org/10.1371/journal.pmed.1001279.