Node-Arc-Node graphs: Acquiring nursing domain expert knowledge Philip Shields RN BaNurse (Hons) PhD candidature nursing Informatics
The problem Is the lack of data describing nursing processes in Donabedian’s quality framework Nurse skill What nurses Patient mix, Culture, do outcomes work load My study addresses the problem by using semantic technology to acquire process data because it paints a picture of human reality in terms of words, that describe concepts and their relationships.
Broad overview of methodology (completed four times) Software robot OWL-DL Semantic & Ontology Logic evaluation Design Science Human Pattern Node-Arc-Node methodology evaluation Graph Usability Study Donabedian ’ s quality framework Nurse Expert
Parts of Semantic Technology The Semantic Web. Made of connected ontologies An Ontology Made of connected triples A Triple Made of three addresses
But, nurses can’t be expected to write a thousand lines of triples to describe their process domain, nurses are not knowledge engineers and knowledge engineers are not nurses. So what’s the answer?
Node-arc-node graphs are the human-readable equivalent of the triple. Graphs are easy to build and connect together with knowledge acquisition software like Visual Understanding Environment (VUE).
Back to the study Methodology Question 1) How can Donabedian’s process domain be evaluated using nurse-constructed node-arc-node graphs? Aim The methodology aims to construct four ontologies from four nurse expert- constructed graphs.
A usability pilot study had a purposive sample of four nursing domain experts with different specialities (Transitional Care, Triage, Surgical, Administration) from the same health cluster. To focus the domain experts on their process domain they were given a ‘starter’ graph containing the three Doran, Dianne., Sidani, Keatings et al. (2002) Nursing Role Effectiveness Model (NREM) process nursing roles: 1) Independent: Autonomous nursing roles 2) Interdependent: Nurses in conjunction with allied health 3) Dependent: Doctor’s order roles The domain experts were instructed to visualise their place in their process environment and ‘fill in the blanks’ from the roles.
The transitional care nurse identified with five links from the independent role, one link from the dependent role and one link from the interdependent role. http://cablesat.com.au/KA/transitional/transitional_nurse.html
The surgical nurse identified with eight links from the independent role, one link from the dependent role and one link from the interdependent role. http://cablesat.com.au/KA/surgical/surgical_nurse.html
The administrative nurse identified with two links from the independent role, no links from the dependent role and one link from the interdependent role. http://cablesat.com.au/KA/administration/administrative_nurse.html
The triage nurse identified with seven links from the independent role, four links from the dependent role and no links from the interdependent role and had no clusters. The patient is shown off to the side because in reality, the patient concept would encompass the entire graph and would lose readability. http://cablesat.com.au/KA/triage/triage_nurse.html
Answers to the methodology Question: How can Donabedian’s process domain be evaluated using nurse-constructed node-arc-node graphs? Graphs, So what? Semantic technology makes the nurse’s contribution to patient care visible. Nurses may use graphs to: 1) Locate ‘edges’ in the graph where embedded technology, indicators or documentation may be placed. 2) Visualise patient care priorities 3) Move resources around to provide better outcomes 4) Graphs provide a ‘bridge’ in which both knowledge engineers and nurses can work together for better patient outcomes. 5) Graphs can be turned into ontologies which are robot readable
Robots, So what? 1) Robots audit the domain structure looking for logic errors. For example, Is a nurse in the right place/has the correct qualifications? Is there a nurse ‘out in the cold’? The start of automated auditing. 2) Robots locate similar semantics across nursing domains for vocabularies and ontology ‘connection points’. Table 2: Term ranking between transitional care and Administrator I-Sub Number Transitional Care Administrator Family Family_Meeting 0.778 Bed_Occupancy_Indicator Bed_Occupancy 0.836 Nurse Nurses .945 ACAS ACAS 1.0 ACAT ACAT 1.0 Care_Plan Care_Plan 1.0 Case_Conference Case_Conference 1.0 External_Agencies External_Agencies 1.0 Hospital Hospital 1.0 IIMS_Reporting IIMS_Reporting 1.0 Policies_and_Procedures Policies_and_Procedures 1.0 WH&S WH&S 1.0 Family Family 1.0
The significance This study provides a window into the process domain as the nurse sees it. Knowledge acquisition techniques such as graphing, semantic classification and logic consistency checking put value on nursing’s contribution to patient care. These techniques bring the nurse's working reality to a stage where it can be visualised and analysed
Limitations 1) Small sample size Nielsen (1994) recommends a sample of not more than five people in a usability study, because he observes, there is no point continuing with a large sample if there is a possibility of an inherent problem in the system being studied. Also, ontologies are completed one at a time and connected together if possible 2 ) No direct way of converting the output of VUE to OWL-DL ontology The graph information had to be transferred by hand from VUE to Protégé which meant printing the graphs and highlighting the node- arc-nodes when they were entered, introducing the possibility of error 3) Am I looking at the process domain? Although the graphs use Doran’s Process roles as a base, the graph may venture into structural and outcome domains
Thank you to my supervisors: Associate Professor Liza Heslop, Dr. Lucy (Sai) Lu and Industry Mentor, Adjunct Professor Evelyn Hovenga.
Donabedian, A. (1988). "The quality of care: How can it be assessed?" Journal of the American Medical Association 260(12): 1743-1748 Doran, D., & Clarke, S. (2011). Toward A National Report Card in Nursing: A Knowledge Synthesis. Eddy, D. M. (1998). Performance measurement: Problems and solutions. Health Affairs, 17(4), 7-25. Goossen, W. (2006). Cross-mapping between three terminologies with the international standard nursing reference terminology model. International Journal of Nursing Terminologies & Classifications, 17(4), 153-164. Grain, H. (2010). Important health Information concepts. In E. J. S. Hovenga, M. R. Kidd, S. Garde & C. H. L. Cossio (Eds.), Health Informatics : An Overview. Amsterdam: IOS Press Naylor, M. D. (2007). Advancing the Science in the Measurement of Health Care Quality Influenced by Nurses. Medical Care Research and Review, 64(2), Suppliment. Needleman, J., Kurtzman, E. T., & Kizer, K. W. (2007). Performance Measurement of Nursing Care: State of the Science and the Current Consensus. Medical Care Research and Review, 64(2), 10s-43s. Pringle, D., & Doran, D. (2003). Patient Outcomes as an Accountability. In D. Doran (Ed.), Nursing-Sensitive Outcomes: State of the Science. London: Jones and Bartlett. Rebstock, M., Fengel, J., & Paulheim, H. (2008). Ontology Engineering Ontologies-Based Business Integration (pp. 97-123): Springer Berlin Heidelberg Sarre, G., & Cooke, J. (2009). Developing indicators for measuring Research Capacity Development in primary care organizations: a consensus approach using a nominal group technique. Health & Social Care in the Community, 17(3), 244-253. doi: 10.1111/j.1365-2524.2008.00821. Van der Bruggen, H., & Groen, M. (1999). Toward an Unequivocal Definition and Classification of Patient Outcomes. International Journal of Nursing Terminologies and Classifications, 10(3), 93-102
Recommend
More recommend
