Informatics principles
| Course home | Appraising information | Keeping up | Informatics principles | Patients | Searching | Wiki help |
Informatics - basic concepts
- Data + knowledge = information
(knowledge is ability to interpet data and turn it into information) - Structured information
- For you - dxplain
- For your patient
Reusing data and the need for automated data aquisition and communication and independence
Example - difficulties with terminology:
A patient has a ABG performed, it shows a PO2 of 45 and a hypercholemic metabolic alkalosis.
Who needs to know about this; what do they need to know and when?
| Who | Why |
|---|---|
| Doc and patient | Clinical care |
| IT staff | Highlight the results in red font to prevent oversight. Create context-sensitive decision support. |
| Billing | the code and whether it supports the instensity of service (eg the doc had to interpret this and spend time responding) claimed |
| Quality Assurance | the result in order to monitor the care given by the hospital to hypoxic patients |
| Medical Researcher | to report on the causes and outcomes of patients with hypoxia |
How to represent thes results of the system successfully understands:
- 'PO2' or 'partial pressure of oxygen'
Solution - controlled vocabularies and terminologies
- About many common clinical abbreviations are in MeSH?
- Multiple meanings of one word: "COLD"
- Try MeSH browser (at PubMed; less known, but offficial version)
- Does MeSH do well at representing a patient's medications and their administration details?
- Mulitple words for one meaning
- How many entry terms for COPD?
- Better?
- Other ways to search
- example is broad topic with multiple meanings hypertension
- Natural language
Example - relationships between terms: semantics / ontologies
The patient has drug adverse effect
- What is dose, preparation. Will MeSH terms represent this? How about NDC?
- Vancomycin level is high - how to convert this is into 'toxicity'
How should a researcher query medical records to retrieve all patients with tuberculosis?
- What terms should be used? What if they only want pulmonary TB?
- What hospital data sources should the researcher look in?
Now, consider how well information resources represent relatioships among facts? Does better representation improve usefullness?
Is uric acid a risk factor for coronary disease?
So what is the chance of coronary disease for: 60 yo male with SPB=130, t chol = 210 and HDL = 40?
- Framingham
- This site gives more than the univariate association of urate and cardiac disease
So is it worth taking statins?
- USPSTF
- This site gives yet more structure to the association between risk factors and cardiac disease
Lastly, even more structure to the knowledge
Other principles
- Malthus law of information - amount of information will outpace human processing ability
