Automated recording of in-patient physiology and the increasing use of electronic health records have created vast data sets that might be harnessed to improve care and used as a substrate for research. Although this data has always been routinely collected, the resolution and scale have significantly improved with automation. By adapting informatics techniques applied in other sciences (e.g. life sciences, economics, sociology, marketing) to analyse data from intensive care patients we hope to identify patterns that indicate association of measured variables with severity of disease and outcomes. This information might be then used as a tool to guide management or for hypothesis generation in future studies.
We aim to create a culture where evidence and data can drive improvement of processes for the benefit of our patients.
Clinical data that is acquired in an autonomous fashion and not as part of a research study mandates that issues with validation, cleaning, missing values and security are addressed. The primary challenge is to successfully detect what are probably small signals amongst the significant amount of noise within intensive care patient data. We hope to be able to generate reproducible workflows that automate the extraction of ‘easy to clean’ data that can be then analysed quickly to give near real-time metrics for measured variables of interest. These metrics are only reliable and valid if the data they are based on is sound and robust.
Delivering improvement to healthcare based on published research or from local data analysis requires adjustment of established behaviours. Challenging embedded processes is the barrier against which most audit or quality improvement often fails. Data is a powerful driver that keeps stakeholders informed about their performance and if used in the correct way, might leverage changes to practice and behaviours.