1. Clinical Guidelines Contextualization in GLARE Alessio Bottrighi*, Paolo Terenziani*, Stefania Montani*, Mauro Torchio #, Gianpaolo Molino # *DI, Univ. del Piemonte Orientale “Amedeo Avogadro” Via Bellini 25/g, 15100 Alessandria, Italy Phone: +39 0131 360174 E-mail: {alessio.bottrighi,terenz,stefania}@mfn.unipmn.it # Az. Ospedaliera S. Giovanni Battista, Via Bramante 88, Torino, Italy - Introduction - GLARE (GuideLine Acquisition, Representation and Execution) - Adaptation to the software context (DBMS) - Adaptation to local (resource) constraints - Conclusions

2. Introduction Many different computer systems managing clinical guidelines (e.g., Asgaard, GEM, Gliff, Guide, PROforma,…) Different roles: -support -critique -evaluation -education -…... Clinical guidelines are a means for specifying the “best” clinical procedures and for standardizing them Adopting (computer-based) clinical guidelines is advantageous

3. Introduction Obstacles to CGL use and dissemination: gap between CGL generality and the peculiarities of contexts of application “General” CGL PRO  general, consensus, standard CONS  difficult application (local constraints) “Local” (e.g., Hospital) CGL PRO  applicability CONS  no generality, no standard, non sharable GLARE’s approach: Semi-automatic adaptation of “general” CGL PRO  general, consensus, standard PRO  applicability

4. Introduction Context Adaptation (1) CONTEXT: LOCAL RESOURCES -Availability of tools for laboratory & test investigations -also: costs and/or time Only CGL actions whose resources are locally available can be executed (2) CONTEXT: LOCAL SOFTWARE ENVIRONMENT (DBMS) -CGL acquisition and execution sw tools must interact with HIS CGL sw tools should be adapted to interact with different DBMS

5. GLARE (GuideLine Acquisition Representation and Execution) -Joint project: Dept. Comp. Sci., Univ. Alessandria (It): P. Terenziani, S.Montani, A.Bottrighi Dept. Comp. Sci., Univ. Torino (It): L.Anselma,G.Correndo Az. Osp. S. Giovanni Battista, Torino (It): G.Molino, M.Torchio -Domain independent (e.g., bladder cancer, reflux esophagitis, heart failure) -Phisician-oriented & User-friendly Some recent pubblications: Terenziani et al., AIIMJ 01,07a,07b, AMIA 00,02,03, Medinfo 04, CGP’04a,04b, AI*IA 03,05, GIN 04,05, AIME 05a,05b,05c AMIA’06 posters T020 and T120 AMIA’06 paper in session S52

6. Representation Formalism Tree of graphs Atomic actions Composite actions (plans) Control relations between actions: -sequence -“controlled” -alternative -repetition (e.g. “3 times each 2 days for a month”) A B C B1B1 B2B2 B 1.1 B 1.2 B 2.1 B 2.2 B 2.3 CG B B1B1 B2B2 D

7. Representation Formalism Hierarchy of Action Types Action Plan Query Work action ConclusionDecision Clinical action Pharmacol. prescription Diagnostic decision Therapeutic decision

8. Representation Formalism description of a clinical action

9. GLARE (main features) Acquisition Tool -Facilities for syntactic & semantic consistency checking -Temporal constraints Execution Tool -Facilities for decision making (hipothetical reasoning facility) -Temporal constraints Interaction with HIS - during acquisition - during execution Testing: bladder cancer, reflux esophagitis, heart failure, ischemic stroke

10. GLARE (Graphical interface)

11. Adaptation to the sw context: GLARE’s three-layered architecture

12. Intermediate XML layer XML as an “interlingua” (legibility, web, dissemination) XML layer and DBMS layer manage the same knowledge, but support different functionalities

13. Three-layered architecture: ADVANTAGES Exploiting an intermediate sharable format (XML) and the DBMS technology to manage large amount of data Making the interaction with users more efficient (Qload+Qstore vs Load+Store) Making GLARE (mostly) independent of the DBMS used for HIS -the interaction of GLARE with DBMS mediated by the XML layer -changing the DBMS  changing only the (Java) module managing the interaction between the XML layer and the DBMS

14. Adaptation to the local context: resources MAIN IDEA: Pre-compilation module INPUT: “general” CGL, list of local resources OUTPUT: CGL context METHOD: CGL context obtained automatically from CGL by pruning non-legal paths (actions) CGL context : adaptation of CGL to the local context: only locally executable paths (actions) are maintained

15. Legal path Def (informal): A legal path is a path in a CGL that contains only locally executable actions (i.e., such that all the resources required by the actions in the path are locally available) Def (inductive): an action belongs to a legal path if (i)it does not require unavailable resources, and (ii)it is the last action of a path, or, alternatively, there is at least one action that follows it which belong to a legal path

16. Algorithm (sketch) All the actions in the input CGL are taken into account. Let A be the current action. 1if A is already present in CGL context (i.e., A belongs also to another legal path in CGL context ), no other actions are required; 2if not, check the set of resources required by A. If A exceeds the available resources then return a failure. Otherwise, A could belong to a legal path; 3 if A is a final action, it is added to CGL context ; 4if A is not a final one, an additional recursive check is applied to every action which follows A in CGL. If there are no legal actions following A, return a failure; TO BE CONTINUED >>>>

17. Algorithm (sketch) 5Take into account the type of A; 6.if A is a composite action, apply recursively the procedure to A’s components. If there is at least one legal path in A: - the set of actions following A is updated with the set of legal actions that follow A previously collected by the recursive calls; - the copy of A (with the adapted internal guideline within) is added to CGL context and returned as a result. otherwise the procedure returns a failure; 7.if the action A is of any other type, no other control is performed. A copy of A is done, and the set of legal actions that follow A is updated and, finally, the copy of A is added to CGL context and returned as a result.

18. Conclusions Context adaptation is crucial to make “general” guidelines applicable in local context In GLARE: automatic adaptation to local resources and semi-automatic adaptation to local DBMS Several related approaches in the literature (e.g., Boxwala, 02; Fridsma et al., 96; Peleg et al., 03; Shahar et al, 96)