Data Analysis in Medical Information Systems
Healthcare providers have entering a new age of co-operation and information technology. In many cases, as in England and Wales, innovation revolves around the creation of shared electronic patient records.
These improvements are leading to more patient data then ever before becomming available to medical decision makers and the health infomaticians who support them.
Many health profesionals and academics would like to see this data be used to aid clinical practice.
The dominant methodology for the teaching and practice of western medicine, Evidence-Based Medicine, asserts that this 'routinely collected' data should only be used to inform practice as an absolute last resort.
DAMIS investigates if this stance is a sound one, or if a need exists to re-examine the assumptions of Evidence-Based Medicine, at least where routine data is concerned.
EBM Asserts: Medical decisions should be based on ‘the best available evidence’ to achieve the best outcome for the patient. Follow this link for a very expansive and widely accepted defition of EBM.
In general the more objective, generalised and quantified a study’s data the better the evidence it is considered to provide. The image below shows a simplified version of EBM's heirarchy of evidence for treatment decisions, follow this link for a more complete version covering all aspects of medical decision making.
Evidence-Based Medicine requires treatment decisions be based on the results of ‘Randomised Controlled Trials’ if available, otherwise the next level of evidence down ‘the hierarchy’ should be used, and so on.
An RCT is a quantitative study to test the relative effectiveness of treatments.
Patients are randomly allocated to either an intervention group or a control group. The intervention group gets a trial treatment, the control group is untreated or treated in the accepted standard way.
Relative effectiveness is judged by comparing the outcomes of the two groups.
EBM assumes that the more objective, scientifically rigorous and quantified the data, the more accurate and justifiable a base it will provide for making decisions. This means that normal patient data collected during real world practice is relegated to the bottom of the evidential heap, to be used as a last resort.
Many within the applied and academic medical communities doubt the relevance of objective, general, quantitative population studies, carried out under controlled conditions on selected patients, to meeting the subjective needs of individual patients in diverse real world clinical practice.
To date there is, ironically, a lack of good evidence that Evidence-Based Medicine is effective at improving patient outcomes.
If the ‘hierarchy of evidence’ theory is sound, past patient data, at the bottom of the hierarchy, should offer no valid information to medical decision makers beyond the information provided by Randomised Controlled Trials, at the top of the hierarchy.
DAMIS adopts this assertion as a null hypothesis and will evaluate it.‘A system based on routinely collected local patient data or using routinely collected local patient data in conjunction with the results of Randomised Controlled Trials will show no greater ability to predict local or individual patient outcomes than analysis of the most relevant Randomised Controlled Trials alone’
The DAMIS project is a collaborative project between
| Cardiff University's School of Computer Science | Velindre NHS Trust |
|
|
| Being carried out by Ben Sissons | |
| Please feel free to contact us with any queries, advice or criticism: b.w.i.sissons@cs.cf.ac.uk |