Machine learning tools for sepsis
Severe bacterial infection or sepsis affects over 5,000 people in Western Australia every year, and around 10% of these will die as a result. Any delay in getting the patient onto the right treatment increases the risk of death, and in regional WA it can be a long time before a patient with sepsis gets the treatment they need. As a result of some exploratory data analysis by a group in the USA, we know that measurable features of blood cells may be able to predict patients with sepsis who are going to do badly.
In a very preliminary data scraping exercise last month we confirmed that these measureable features are routinely collected for all patients whose blood is tested at Sir Charles Gairdner Hospital. That data is archived but not included in pathology reports. The raw data is available in CSV format in Excel spreadsheets but needs a lot of cleaning, and its alignment with other critical data is highly time consuming. Once scraped and cleaned, the blood data can be used for exploratory data analysis, visualisation and predictive modelling in Orange or similar.
What we urgently need is help with data scraping and cleaning with Excel files, development of suitable macros to assemble the data for subsequent analysis in Orange, and optimisation of the machine learning pipeline so that we can work out the best method for early identification of the paients with suspected sepsis who are going to do badly without more active treatment.
References
-
Time to Treatment and Mortality during Mandated Emergency Care for Sepsis. Seymour CW, Gesten F, Prescott HC, Friedrich ME, Iwashyna TJ, Phillips GS, Lemeshow S, Osborn T, Terry KM, Levy MM. N Engl J Med. 2017 Jun 8;376(23):2235-2244. doi: 10.1056/NEJMoa1703058.
-
Improved Early Detection of Sepsis in the ED With a Novel Monocyte Distribution Width Biomarker. Crouser ED, Parrillo JE, Seymour C, Angus DC, Bicking K, Tejidor L, Magari R, Careaga D, Williams J, Closser DR, Samoszuk M, Herren L, Robart E, Chaves F. Chest. 2017 Sep;152(3):518-526. doi: 10.1016/j.chest.2017.05.039.
Field trials of blood culture identification FilmArray in regional Australian hospitals.
Bzdyl NM, Urosevic N, Payne B, Brockenshire R, McIntyre M, Leung MJ, Weaire-Buchanan G, Geelhoed E, Inglis TJJ. J Med Microbiol. 2018 Mar 13. doi: 10.1099/jmm.0.000714.
Client
Contact Person: Tim Inglis, Pathology & Laboratory Medicine, School of Medicine, UWA
Telephone: 0407 94 631
Email: [email protected]
Preferred method of contact: e-mail
Location: QEII Medical Centre, office in PathWest building, lab in L block Microbiology
Client Unavailability
None
IP Exploitation Model
The client wishes to use a Creative Commons CC BY-NC model to deal with IP embodied in the project.