Antibiotic treatment for various infections is typically started without knowing the bacteria causing the infection or which antibiotics they are susceptible to. To avoid delays in effective treatment, the initial treatment choice should try to avoid antibiotics the bacteria have a high probability of being resistant to. This aim must be balanced against the need to avoid prescribing very broad-spectrum or (near) last resort antibiotics unnecessarily, in order to reduce the risk of antibiotic resistance.
Increasing availability of large national and local linkable datasets provides an important opportunity to enable better-informed empirical antibiotic treatment decisions. Statistical prediction and machine learning models can be applied to these datasets to predict which antibiotics bacteria are likely to be resistant to. By incorporating information about, for example, patient demographics, prior hospitalisations, prior antibiotic use, whether the infection is hospital-acquired, and accounting for interactions between these factors we will provide more accurate predictions of appropriate empirical antibiotic therapy than current approaches. The project will be necessarily cross disciplinary, requiring state-of-the-art statistical and machine learning techniques to gain insight from ‘big data’, with crucial clinical and biological input needed around the complexities of the patient pathways/epidemiology of antibiotic resistant bacteria. Thus ensuring the models developed are clinically relevant with potential to improve health policy.
Such predictions could lead to substantial improvements in the chances of choosing appropriate antibiotics via personalised empirical treatment, while minimizing the risk of further increasing resistance.
The potential impact in the NHS of implementing such models aiding prescribing decisions on patient outcomes, development and spread of antibiotic resistance, and cost-savings will be assessed using computer simulations.
The student will develop and improve skills in statistical, machine learning, and health economic modelling. As this is an interdisciplinary project it is anticipated that the student may have academic skills and experience in, for example, a computational or an epidemiological discipline but not both. Dependent on the academic background of the student, we will provide informal - or where required and as funds permit formal - training in microbiology, statistical and/or machine learning modelling, or health economic modelling as appropriate. The supervisory team is multi-disciplinary with both technical and clinical expertise in antimicrobial resistance, mathematical, statistical and economic modelling, infectious disease epidemiology and intervention evaluation.
Project reference number: 1024
|Dr David Eyre||Experimental Medicine Division||Oxford University, John Radcliffe Hospital||GBRemail@example.com|
|Dr Julie Robotham||Public Health England, Colindale||GBR|
|Dr Koen Pouwels||Public Health England||GBR|
|Professor (Ann) Sarah Walker||Experimental Medicine Division||Oxford University, John Radcliffe Hospital||GBRfirstname.lastname@example.org|
|Professor Ben Cooper||Tropical Medicine||Oxford University, Bangkok||THA||Ben@tropmedres.ac|
There are no publications listed for this DPhil project.