Knowledge-based decision support for patient monitoring in cardioanesthesia

A knowledge authoring tool for clinical decision support

Anesthesiologists in the operating room are unable to constantly monitor all data generated by physiological monitors. They are further distracted by clinical and educational tasks. An expert system would ideally provide assistance to the anesthesiologist in this data-rich environment. Clinical monitoring expert systems have not been widely adopted, as traditional methods of knowledge encoding require both expert medical and programming skills, making knowledge acquisition difficult. A software application was developed for use as a knowledge authoring tool for physiological monitoring. This application enables clinicians to create knowledge rules without the need of a knowledge engineer or programmer. These rules are designed to provide clinical diagnosis, explanations and treatment advice for optimal patient care to the clinician in real time. By intelligently combining data from physiological monitors and demographical data sources the expert system can use these rules to assist in monitoring the patient. The knowledge authoring process is simplified by limiting connective relationships between rules. The application is designed to allow open collaboration between communities of clinicians to build a library of rules for clinical use. This design provides clinicians with a system for parameter surveillance and expert advice with a transparent pathway of reasoning. A usability evaluation demonstrated that anesthesiologists can rapidly develop useful rules for use in a predefined clinical scenario.

Evaluation of two new ecological interface approaches for the anesthesia workplace

OBJECTIVE

Currently, vital parameters are commonly displayed as trends along a timeline. However, clinical decisions are more often based upon concepts, such as the depth of anesthesia, that are derived by combining parameter relationships and additional context information. The current displays do not visualize such concepts and therefore do not optimally support the decision process. A new display should present an ecological interface (EI). The principle of EI design is to visualize all of the information necessary for decision making in one single display.

METHODS

In the first approach, we developed an EI that visualizes 35 relevant parameters for anesthesia monitoring. All of the parameters are generated by an anesthesia software simulator. Sixteen anesthetists had to administer two simulated general anesthetics: in one setting working only with the simulator's monitors ("Sim Only"), and in another setting working with the simulator's monitors in combination with the EI ("Combi1"). During each experiment, one unexpected critical incident (either blood loss or a cuff leakage) had to be identified. The control and monitoring behavior was analyzed by recording the subjects' eye movements and think-aloud protocol. With the help of the eye-tracking results, we re-designed the EI. The new EI was then tested with no eye tracking ("Combi2") on eight anesthetists under analogous conditions as in "Combi1."

RESULTS

Cuff leakage was identified significantly quicker in "Combi1" (7 of 8 cases; time (T): 65 s +/- 73 s) than in "SimOnly" (6 of 8 cases; T: 222 s +/- 187 s). Blood loss was identified in 5 of 8 cases (T: 215 s +/- 76 s) in "Combi1" as quickly as in "SimOnly" (all cases; T: 217 s +/- 72 s). In "Combi1," the EI was used as the main source of information (in 43 +/- 19% of time) and was frequently favored when identifying an evolving critical incident. In "Combi2," cuff leakage was identified in 7 of 8 cases (T: 70 s +/- 111 s) as quickly as in "Combi1." Blood loss was identified significantly quicker in all cases (T: 147 s +/- 62 s) in "Combi2" than in "Combi1" and in "SimOnly."

CONCLUSION

The results have shown that appropriately designed EIs may improve the anesthetist's decision making and focus attention on specific problems. Now, the findings have to be tested in future studies by widening the scope using other simulated scenarios and being closer to reality under real conditions in the OR. Eye tracking proved to be a useful method to analyze the anesthetists' decision making and appropriately re-design interfaces.

Automated information systems in anesthesiology

The current state of the art of anesthesia information systems remains primitive. Currently, available commercial systems focus only at automating the charting process and not the care process. Until systems are available that integrate these two functions, anesthesiologists will not truly benefit from such systems.

Development of a decision support system to assist anesthesiologists in operating room

The complexity of modern anesthesia procedures requires the development of decision-support systems functioning in a smart-alarm capacity. We developed computer algorithms to detect critical conditions during surgery (light anesthesia or unstable blood pressure), based on computerized anesthesia records containing hemodynamic data (heart rate, mean arterial pressure and systolic arterial pressure). Our analysis indicated that a > or = 12% change in mean arterial blood pressure (MAP), compared with the median value of MAP over the preceding 10-min interval, may be chosen as the criterion for detecting LA, with a sensitivity of 96% and a specificity of 91%. The best agreement between human and computer ratings of blood pressure lability (correlation coefficient 0.78) was achieved when we used the absolute value of the fractional change of the mean arterial pressure (magnitude of FCM) between one 2-min epoch and the next 2-min epoch. Work is under progress to develop a decision-support system to alert clinicians in the operating room environment to critical events.

Ergonomic evaluation of an ecological interface and a profilogram display for hemodynamic monitoring

OBJECTIVE

Comprehensive monitoring of the patient state and subsequent decision making is an essential part of the task of an anaesthetist. The physicians' decision making process is based upon a concept of partly abstract physiologic parameters such as depth of anaesthesia or contractility. This concept is derived from the measured parameters given on todays' trend displays in addition to context information available for the anaesthetist. We investigated two alternative approaches of display design for hemodynamic monitoring: 1) integrated displays based on ecological interface design, and 2) profilogram displays based on intelligent alarms.

METHOD

To evaluate differences in decision making, the two displays and a trend display were compared in an experimental set-up with computer simulated vital parameter curves. From a start state with random parameter deviations from the ideal state, subjects had to achieve the ideal circulatory performance as fast as possible by manipulating vasomotor tone, heart rate, blood volume and contractility. To analyse subjects' decision making process, eye-tracking, event-logging, and the method of think aloud protocols were used. Twenty anaesthesiologists performed 113 experiments (approximately 2 with each display).

RESULTS

The anaesthetists failed to achieve the task in 37% using the trend display, in 19% using the profilogram display, and in 13% using the ecological interface. Hence, a safer task solution was possible with the ecological interface and the profilogram display but at the expense of various performance parameters such as higher trial time, more interactions with the simulated system, and more frequent eye movements. In contrast to the trend display and the profilogram display, where anaesthetists were mainly focussed on controlling the left atrial pressure, such an behaviour was less observed with the ecological interface.

CONCLUSION

Our results have shown that subjects came to more effective solutions with the traditional trend display. The main reason for this result may be their years of experience with this kind of display type. Regarding safe and goal-intended decision finding, the results are encouraging for further experiments with redesigned ecological displays. But these displays ought to have smoother changes with respect to the traditional trend displays. Furthermore, new experiments have to be performed under real or fairly real (e.g. together with an anaesthesia simulator) conditions to underline the positive results for ecological interfaces.

Ergonomics applied to anaesthesia record keeping

For almost 100 years, the anaesthesia record has been the sole information tool trying to fulfill an ample catalogue of functions related to the anaesthesia information processes. Automated anaesthetic record systems have evolved around data being available online, as an imitation of the handwritten record. None has developed an information tool capable of an efficient utilization of the wide range of resources provided by modern technology to fulfill the information requirements of the anaesthetic environment. We used a system ergonomic analysis trying to find the best solutions. As a result of it we drafted an Anaesthesia Information Concept (AIC) in which the complexity of data & information (D&I) processes is broken down to modules called Clinical Information Process Units (CIPUs). A CIPU is mainly defined by the responsibility of a staff member and focuses on the basic system patient, staff and machine (all devices). The internal functions of a CIPU are treatment control and medicolegal documentation. The external functions are fulfilled by transferring required sets of D&I for subsequent treatment control (next CIPU), audit, quality control, cost calculation, etc. Using such an approach, an Anaesthesia Information Concept (AIC) can be realized by a wide range of modular and hybrid systems (combination of different tools such as paper records, computers, etc), as opposed to universal and single automated documentation systems, which up to now have failed to fulfill the information demands of the anaesthetic environment.

Left ventricle automated detection method in gated isotopic ventriculography using fuzzy clustering

A method that uses the fuzzy ISODATA clustering algorithm and Fourier analysis is proposed for automated detection of heart left ventricle contours. This operation is used for quantitative analysis of cardiac function. The computation begins by finding the phase image. The fuzzy ISODATA algorithm is first applied to this image to generate a number of clusters that correspond to the organ substructures (ventricles, atria). Second, the ventricles cluster is isolated and the intensities of its points are replaced by the corresponding ones from the original (end diastolic) frame. Finally, a reduced image representing the ventricular region is obtained and an additional clustering is performed to find the left ventricular boundary automatically. This algorithm is tested by application of 105 sets of 16 images each. These results are compared with the measurements obtained with two semi-automatic methods used, respectively, on the Philips and the Sopha Medical gamma cameras.