Accessibility Skip to Global Navigation Skip to Local Navigation Skip to Content Skip to Search Skip to Site Map Menu

Diagnosing ventilator-associated pneumonia before it happens

A 2018/2019 Summer Studentship research project

This study will quantify and delineate the association between VAP and insulin sensitivity in critically ill patients. If this association is strong, as recent published results indicate, it would create a tool to predict the onset of VAP before it occurs, allowing better oversight and ventilation management, as well as improving outcomes and reducing costs.

Student: Ellen O’Byrne
Supervisor: Professor G M Shaw, Dr J K Dickson, Professor Geoffrey Chase
Sponsor: TBC

Project brief


Ventilator-associated pneumonia (VAP) is a major cause of morbidity, mortality, and cost. Recent published studies show a spike in hyperglycemia caused by reduced potential insulin sensitivity in the 24-48 hours preceding VAP. This discovery opens the possibility to employ our model-based insulin sensitivity metric (SI) used in our STAR glycemic control (GC) protocol, which is the standard of care in the ICU here and elsewhere internationally. This metric is highly validated and thus could be correlated to diagnosed cases of VAP in our ICU to assess if there is a strong, diagnostic and predictive association between the occurrence of VAP and the trends in SI. If so, it would create a novel new predictor and diagnostic.


To assess the relationship between validated SI and VAP in retrospective clinical data, and create a predictive diagnostic.


All STAR data is cloud stored and there is existing ethics approval to audit this data for research and safety purposes (URA/12/EXP/004). Patients (N = 500) from STAR will be evaluated to find those patients whose diagnosis of pneumonia meets the criteria for VAP. There insulin sensitivity will be computed at the University of Canterbury using our proprietary models and methods. These data will then be used to create basic diagnostic statistical models. Data for all patients includes:

  • Age and Sex
  • APACHE II and III score
  • Organ failure scores and data (SOFA score)
  • Primary / Secondary diagnoses
  • Blood glucose data
  • Insulin given
  • Enteral and Parenteral nutrition (EN and PN)
  • ICU and Hospital Mortality
  • Length of Stay
  • Length of time on GC

Main outcomes

Results will model the association between VAP occurrence and (prior) insulin sensitivity (SI) patterns in critically ill patients. This model will be assessed for its ability to predict and diagnose impending VAP, including how far ahead in time it can make this prediction.


The student will be exposed to methods and practice of basic research and clinical practice. He/she will be involved in the collection of data, study analysis, and refining the study analysis/design. They will thus have experience of a wide range of clinical data and its analysis for assessing safety and quality. Their work forms part of a larger research programme carried out by senior academic staff (Dist Prof JG Chase, Dr JL Dickson), PhD students from the University of Canterbury, and Prof GM Shaw from ICU. Results from this work will be published in appropriate medical and biomedical engineering journals.


Ideally the student will be involved in all clinical data gathering and analysis. The student will have supervised, hands-on, experience in data collection and initial clinical analysis (5 weeks). This will be followed by statistical modeling analysis to determine the associations between VAP and SI (3 weeks). Documentation of results at the end (2 weeks).