Automatically detecting interactions between medication rules for elderly patients
About Automatically detecting interactions between medication rules for elderly patients
- Contact person: has supervisor::Annette ten Teije
- This project has not yet been fulfilled.
- This project fits in the following Bachelor programs: {{#arraymap:|, |xXx|bachelorproject within::xXx|,}}
- This project fits in the following masterareas: {{#arraymap:Human Ambience, Technical Artificial Intelligence, Knowledge Technology and Intelligent Internet Applications|, |xXx|project within::xXx|,}}
Description
GOAL Interactions between guidelines is an increasing problem in a increasingly elderly population, with increasing prevalence of comorbidities. It is currently difficult to detect such interactions because guidelines are written independently from each other.
In this project, we will test a method for semi-automatic detection of guideline-interactions in the domain of medication for elderly patients.
MATERIALS
- ACOVE-NLI is a set of 392 clinical quality indicators for elderly care. These quality indicators can also be read pro-actively as decision-support rules. http://www.rand.org/content/dam/rand/www/external/health/projects/acove/docs/acove_qimedadmin.pdf
- MedLock@KIK has formalised 40 of these rules in if-then format, in particular ruleson avoiding inappropriate medication (as well as rules on continuity and documentation of care). http://kik.amc.uva.nl/home/aabuhanna/LERMapplication.pdf
- Veruska Zamborlini (KRR@VU) has developed a light-weight formal method to detect interactions between multiple guidelines. http://www.mendeley.com/c/7238622494/p/22486701/zamborlini-2014-a-conceptual-model-for-detecting-interactions-among-medical-recommendations-in-clinical-guidelines/
RESULTS Measures of precision and (if possible) recall on identified interactions between medical decision rules on medication for elderly patients.
METHOD
1. reformulate MedLock's formalisation of the ACOVE-NLI subset inZamborlini's TMR-i model
2. implement the resulting model in computer-executabe form
3. use the resulting implementation to infer possible guideline interactions
4. validate the detected interactions with a domain expert on soundness (precision) and completeness (recall).
STAFF INVOLVED
- MedLock from KIK,
- Veruska Zamborlini & Annette ten Teije from KRR@VU,
- Andrea Maier from VUmc.
