Developing automated risk stratification for infectious complications in secondary immunodeficiency - Promotieonderzoek

Diseases that used to be very debilitating, such as rheumatoid arthritis (RA), multiple sclerosis (MS), or inflammatory bowel disease (IBD), or even fatal, like cancer, can now be treated successfully with drugs that target the immune system. These treatments, however, can induce secondary immunodeficiency (SID) as a side effect, which increases the risk for potentially fatal infections. That interplay is particularly challenging for elderly patients. Many healthcare professionals are insufficiently acquainted with the various forms of SID that these treatments can induce and the associated infectious risks for the patients. As a consequence, patients may develop infections that could have been prevented or at least detected at an earlier stage, if the treating physician had been alerted to the risks. These so-called ‘opportunistic’ infections cause additional morbidity and mortality, increased numbers of hospital admissions and higher prescription numbers of antibiotics. In sum, the opportunistic infections add substantial costs to the healthcare system and seriously affect the patients’ recovery and prognosis.

The project’s objectives are to i) harness techniques from natural language processing to extract and quantify data from free-text fields in Epic®, ii) apply techniques from supervised machine learning to learn from historical patient data to predict infection risk, and iii) examine how a human-in-the-loop approach can improve the predictive performance and the healthcare professionals’ acceptance of such a tool in the decision-making process.