[ Pobierz całość w formacie PDF ]
.It is significant that inmodeling we found that we could conceive of a general static model, but thebehavioral part was much more dependent on individual departments.Of course, astatic model implies a great deal of behavior.Behaviors exist to createobservations and to provide various ways of navigating associations to understandhow those observations fit with other observations.The behavior we cannot imply,however, is the sequence of observations that a typical department makes.Often aclinician has some path of observations that can be taken.Departmental policymay be to record this path in terms of higher-level protocols (see Chapter 8).It isdifficult, and almost certainly impossible, to design a general process that allclinicians could use.It is possible, however, to sketch an outline of the process involved in makingobservations.I begin by looking at how making a new observation can triggerfurther observations, as shown in Figure 3.15.Whenever clini cians makeobservations, they consider the possibility of other associated observations.Theyuse the associative functions they know to come up with a list of possibleobservation concepts that might be associated with the triggering observations.They can then propose further observations as needed.52 Process of ObservationFigure 3.15 Making an observation triggers further observations.Further observations are suggested by the knowledge level.In Figure 3.15 the concurrent trigger rule is labeled "associated observationconcepts." In event diagrams, trigger rules have two purposes.First, they showcause and effect.When we are considering business processes, this is usuallyenough, but as we delve deeper we see a second purpose.Any operation has inputand output.The trigger that connects two operations must describe how to getfrom the output of the triggering operation to the input of the triggered operation.In many cases this is trivial, as they are the same object (as in the trigger frompropose observation to make observation shown in Figure 3.15).However theycan get quite complex, as in finding associated observation concepts.When we have a more complex trigger rule, we can represent the trigger rulewith another event diagram.Figures 3.16 and 3.17 do this for the associatedobservation concepts trigger.We begin by finding all the associative functionswhose input includes the initial observation's observation concept.We thenevaluate each of these associative functions.For each one that evaluates to true, wefind the product and add it to the answer.Since these event diagrams describe atrigger rule query, all the operations must be accessors and hence must not changethe observable state of any object.Figure 3.16 Event diagram to describe the query for finding associated observations.This lies on the concurrent trigger of Figure 3.15 or in the operation of Figure 3.18.When the trigger rule query is complex, you can also represent the query as anoperation in its own right, as shown in Figure 3.18.Either method is correct.Observations and Measurements 53Figure 3.17 An interaction diagram for finding the possible observation concepts impliedby an observation.This interaction diagram supports Figure 3.16.Figure 3.18 Notating the query explicitly as an operation.This is equivalent to Figure 3.15.You can either show queries as operations or considerthem part of the trigger, trading simplicity for compactness.Even after the query, there is a control condition (evaluate proposal) beforean observation is proposed.The query suggests possible observation concepts tolook for based on the associative functions.This step could easily be done bysoftware in a decision support system.The control condition represents the extrastep of deciding whether the suggested observation concept is worth testing for.We did not feel we could formally model this process, implying that this step isbeyond software and can only be done in the clinician's head.Figure 3.19 includes additional triggers that arise from projections andactive observations.The triggers to propose intervention work in a similar wayto the previous case.We suggest interventions that are eval uated by theclinician before they are proposed.This reinforces the fact that although anyobservation can lead to further observations being made, only active observationsor projections (not hypotheses) lead to interventions.(An intervention54 Process of Observationis an action which either intends or risks a change in state of the patient [ Pobierz całość w formacie PDF ]