ESCTAIC Abstracts ESCTAIC AND SCATA: AUTUMN MEETING

November 11^13, 1999 HCI International Medical Centre, Glasgow

HAND HELD RECORDS J. M. Watt,1 J. A. Lack,2 1 Featherstone Department of Anaesthesia, University Hospital, Birmingham, 2 Department of Anaesthesia, Salisbury Hospital, Salisbury An anaesthetic record database program using the Psion series II was used by all anaesthetists at the General Hospital Birmingham from 1990 until the hospital's closure in 1994. In this period over 11000 anaesthetic records were made using the system and individual logbook data was produced as well as a monthly departmental audit. In 1994 the Psion 3a was introduced with the important advantages of a full QWERTY keyboard and 4 line display; although a more conventional PC database was used in the main theatres, a roving terminal has proved useful for record keeping in areas without such equipment. The RCA/SCATA guidelines for anaesthetic records were used in the construction of a new database which employed a hierarchical menu structure to select entries from a series of pick lists. Pre-operative, intra-operative and post-operative data could be entered at di¡erent times allowing continuity of the record between the ward, theatre and recovery room. The data could be transferred to a PC for storage and processing. Practical use of the system by one author (JW) on 478 cases demonstrated that it was a useful supplement to the InTheatre PC based record keeping database. Its main advantage is portability for pre and post-operative visiting as well as anaesthesia in ``remote locations'' (x-ray, resuscitation room, A&E etc.) Data from these cover an interesting period in the development of anaesthesia, with the introduction of Propofol and the laryngeal mask amongst other changes REFERENCES 1. Lack JA, Stuart-Taylor M, Techlenburg A. An anaesthetic minimum dataset and report format. BJA 1994; 73 (2): 256^260 2. Anaesthetic Record Set Royal College of Anaesthetists Newsletter, April 1996 STANDARDISATION OF JUNIOR DOCTORS' PROTOCOLS M.Y. Vanarase, J. A. Lack, Salisbury District Hospital,Wiltshire Introduction. Many anaesthetic departments have devised protocols for trainees to introduce them to the routines of the hospital and for appropriate management of various clinical situations. However there is no consistency about either their subjects or style. Aim. To survey current practice regarding junior doctors' protocols and to design a standardised framework of clinical and non-clinical guidelines, which can be easily accessed. Materials and Methods. We wrote to clinical tutors in all the Anaesthetic departments of the Wessex region requesting them to send us any protocols they had. A follow up phone call was made as a reminder. Protocols from di¡erent Hospitals were scrutinised by two independent observers and categorised into di¡erent clinical and non-clinical situations. Similarly

Journal of Clinical Monitoring and Computing 16: 141^158, 2000. ß 2000 Kluwer Academic Publishers. Printed in the Netherlands.

142

Journal of Clinical Monitoring and Computing Vol 16 No 2 2000

algorithms and action plans published by various sources e.g. AAGBI, Resuscitation council etc. were studied and relevant points included in the ¢nal framework. Results. Major section headings were decided and good points from each protocol were selected & formatting of the protocols was done. The major sections are ^ Induction programme, Guidelines for a new SHO, On calls, Clinical protocols & Action plans, Pain management, Obstetrics & Paediatric protocols, ITU & Day unit protocols. Each protocol has a section to be completed to describe local practice e.g. where things are kept, call out routines etc. The framework will be available on Internet. We intend to introduce this framework on HIS or as a booklet in each Hospital thereby ensuring easy access. Discussion. This is an attempt to derive ``best of breed'' set of protocols without trying to be critical of those which are currently used. The suggested framework may be modi¢ed locally to meet individual department's needs. Easy access and regular updating will make it a good reference guide for each anaesthetic department. A CLINICAL REPORTS ``WIZARD'' FOR THEATRE Dr William A. McCulloch, Consultant Anaesthetist, Hairmyres Hospital, East Kilbride Introduction. Despite the collection of large quantities of theatre data in this District General Hospital over the last 6 years, it has been di¤cult to give clinicians the information they need to improve the quality of their service. A desktop tool has been designed to meet this need through a reporting Wizard that is simple to install and use, and can generate reports, graphs or data ¢les on any subset of theatre information. Methods. From a technological point of view, the aims of the project were as follows. a) The development environment was to be Delphi 3. No none-native-Delphi products were to be used - for simplicity and to minimise installation problems. b) Object Orientated techniquess were to be used as widely as possible so that we could learn their place in projects of this type. c) Data acquisition was to be separate from data analysis , in order to allow virtually seamless analysis of data from two completely di¡erent database systems and structures (Paradox for DOS, and Oracle on UNIX) - ``database independence''. The system collects initial parameters from the user, indicating whether the report is for an Anaesthetist or Surgeon, whether it is to be specialty, session, surgeon or operation-type based, and what period the report is to cover. In general, theatre system users are allowed access to details of patients they have ``owned'', operated on, or anaesthetised, although specialties can allow unrestricted access by any member of that specialty. An initial sweep of data collects the information for each from the database into a descendent of a ``tPatientDetails'' object, so that anaesthetic reports collect di¡erent information (ASA, agent used, anaesthetic type, and critical incidents) from surgical reports, although the ``objects'' have a large common core. On a low spec Personal Computers (P120, 16

MB) the system can handle in excess of 10,000 patient records in memory, although it may take 10 minutes to gather these. The data collected from the theatre system is then related to data held on other ``friend'' systems, such as the list of deaths from the Patient Administration System and the Postoperative wound infection audit, to enrich the dataset. Once the data has been collected, it is easy to query the information which is now held in memory. Instantaneous ¢lters can be easily applied and removed on criteria such as postoperative death, cases after midnight, age, critical incidents, duration of surgery. Filters can be applied on top of each other to any level. The system will then print the data as a log, listing details of patients in the current ¢ltered group, graph the data to demonstrate distribution by parameter, or export the data to a text ¢le for further analyis. Discussion. The above approach is now in use within our hospital and is beginning to allow clinicians the data they need to design and optimise services, as well as straightforward data like logbooks for all junior sta¡. Further developments will require the acquisition of better outcome data (postoperative complications, pain scores etc). We believe that the ability to browse and ¢lter data in this way greatly contributes to the clinical usefulness of our theatre system. DATA CIRCULATION AT ANAESTHESIOLOGICAL WORKPLACES USING AN ANAESTHESIA INFORMATION MANAGEMENT SYSTEM (AIMS) L. Quinzio, A. Junger, M. Benson, G. Sciuk, A. Jost, G. Hempelmann, Department of Anaesthesiology and Intensive Care Medicine, Justus-Liebig-University Giessen, Germany Objective. The anaesthesia information management system (AIMS) at the department of anaesthesiology and intensive care medicine of the Justus-Liebig-University of Giessen registers the data of any anaesthesia procedure decentralised at every work place in an automated record. This study intends to show if the system in its current con¢guration is suitable to provide data for the quality documentation and to ensure a feedback on the performed anaesthesia work. This feedback is achieved by presenting the anaesthesia data on request at each workstation after a centralised evaluation and statistical processing. Methods. The data sets of all anaesthesia procedures since 1997 have been recorded with the anaesthesia documentation software NarkoData 4 (ProLogic GmbH, Erkrath, Germany), archived in a relational Oracle 7 database (Oracle Corporation, Redwood Shores, USA) and processed statistically using the SQL-based program Voyant (Brossco Systems, Espoo, Finland) [1]. Selected statistics are made available on the hospital's intranet via a mail server dependent on user access privileges. The statistics are updated periodically and can be accessed with a standard HTML-browser. Results. Based on the data sets of the 41,393 anaesthesia procedures recorded in 1997 and 1998 (database status at 01/ 04/1999), performance statistics such as anaesthesia time or number of anaesthesia procedures can be accessed at each of the currently 131 local AIMS workstations. These statistics can be modi¢ed according to the users preferences, selecting views depending on various criteria such as department, time period and operating room. Furthermore, information for

ESCTAIC Abstracts 143

internal quality management [2] is available, for example statistics on adverse events according to type, operation (ICPM) and ASA classi¢cation. Conclusion. On the basis of the large number of complete and valid data sets, an AIMS produces powerful statistics [3]. With the presentation at each anaesthesia workstation, the circle going from the recording of the data to the data evaluation and back to the workplace is closed for the documenting person. This is a decisive step for an e¡ective feedback in an information management system. Especially for quality documentation this immediate presentation of the results could lead to an improvement of quality. REFERENCES 1. Benson M, Junger A, Quinzio L, Michel A, Marquardt K, Hempelmann G. Erfahrungsbericht u«ber drei Jahre Routinebetrieb eines Ana« sthesie-Informations-Management-Systems (AIMS) am Universita« tsklinikum Giessen. Ana« sthesiol Intensivmed Notfallmed Schmerzther 1999; 34: 17^23 2. DGAI-Kommission ``Qualita«tssicherung und Datenverarbeitung in der Ana« sthesie'': Empfehlungen zur Nomenklatur und Erfassung von Daten zur Qualita« tssicherung. Ana« sth Intensivmed 1995; 36: 72 3. Edsall DW, Jones BR, Smith NT. The anesthesia database, the automated record, and the quality assurance process. Int Anesthesiol Clin 1992; 30: 71^92 INFLUENCE OF DIFFERENT RECORDING METHODS ON DOCUMENTATION WITH AN ANAESTHESIA INFORMATION MANAGEMENT SYSTEM A. Junger, M. Benson, L. Quinzio, C. Fuchs, G. Hempelmann, Dept. of Anaesthesiology and Intensive Care Medicine, Justus-Liebig-University Giessen, Germany Background. In 1997 and 1998, the data of 41,393 anaesthesiological procedures have been recorded online with an anaesthesia information management system (AIMS) and imported into a database at the department of Anaesthesiology and Intensive Care Medicine of Justus-Liebig-University Giessen. The objective of this study was the evaluation of the in£uence that the method of recording in relation to the degree of automation had on the quality of documentation during the procedure. Methods. Firstly the variability of di¡erent vital parameters (SBP, DBP, MAP, HR, SpO2 , Temp) in relationship to the mode of documentation (manual or automated) has been evaluated for two separate ASA classes (ASA II). The procedure was subdivided into the following intervals: pre-, intra- and post-operative. In relation to the total number of measurements the mean value and the variance of all measurements per patient have been determined for each interval. With these ``individual'' average values and variants an overall mean value was calculated for each patient in the categories ``manual data gathering'' and ``automated data gathering.'' In order to check the hypothesis ``similar variance in both collectives,'' the non-parametric ``Wilcoxon ranking sum test'' has been carried out. For the adverse event ``hypotension'' the di¡erent number of incidences in relation to the mode of recording has been evaluated: 1. manual recording of adverse

Mode of recording

Frequency

Manual recording of adverse events (DGAI)

352 (1.8%)

Computerised automatic detection of manually entered blood pressure values (no online data transfer from vital monitor)

210 (3.6%)

Computerised automatic detection of blood pressure values with online transfer of vital data

1429 (10.4%)

Automatic detection total

1639 (8.5%)

events according to the regulations of the German Society for Anaesthesiology and Intensive Care Medicine (DGAI) [1], 2. computer based automatic detection of manually entered values for blood pressure (without online data transfer), 3. computerised automatic detection of online recorded blood pressure values. For automatic detection the adverse event ``hypotension'' has been de¢ned as decrease of mean arterial blood pressure (MAP) of more than 30% during 10 minutes and the administration of a vasoconstrictor until 20 minutes after onset of decrease (volume administration has not been considered). Results. The variation of automatically recorded vital parameters SBP, DBP, MAP, HR and SpO2 was signi¢cantly higher during all three intervals and in each ASA group than the variation of manually recorded measurements. For the body temperature this was true only during the immediate postoperative period. The frequency of all patients with at least one adverse event ``hypotension'' in relation to the mode of recording is 1998 (n = 19,249) in shown in the table below. Discussion. The ¢ndings support the assumption that anaesthesiologists tend to document a ``smoothed trend.'' Similar to other studies [3] we also ¢nd a discrepancy between manually recorded data and data supplied by the vital signs monitor. The increasing degree of automation with data collection and evaluation identi¢es an increasing number of adverse events such as ``hypotension.'' An AIMS with automated data transfer allows an exact documentation in time especially during induction and, at the end of an anesthesia [2]. REFERENCES 1. DGAI-Kommission ``Qualita«tssicherung und Datenverarbeitung in der Ana« sthesie'': Empfehlungen zur Nomenklatur und Erfassung von Daten zur Qualita« tssicherung. Ana« sth Intensivmed 1995; 36: 72 2. Edsall DW, Deshane P, Giles C, Dick D, Sloan B, Farrow J. Computerized patient anesthesia records: less time and better quality than manually produced anesthesia records. J Clin Anesth 1993; 5: 275-283 3. Sanborn KV, Castro J, Kuroda M, Thys DM. Detection of intraoperative incidents by electronic scanning of computerized anesthesia records. Comparison with voluntary reporting. Anesthesiology 1996; 85: 977^987

144

Journal of Clinical Monitoring and Computing Vol 16 No 2 2000

EVALUATION OF AN ECOLOGICAL INTERFACE FOR THE ANAESTHESIA WORKPLACE BY EYE-TRACKING A. Jungk,1 B.Thull,1 A. Hoeft,2 G. Rau,1 1 Helmholtz-Institute for Biomedical Engineering, Aachen University of Technology, Germany; 2 Department of Anaesthesiology and Intensive-Care-Medicine, Bonn, Germany Introduction. On today's displays vital parameters are commonly displayed as trends along a timeline. However, the anaesthetists' decision making process is based upon a concept of partly abstract physiologic parameters (e.g. depth of anaesthesia) which is derived by combining parameter relationships and additional context information. The current displays do not visualise this concept and do therefore not optimally support the decision process. Our preliminary studies [1, 2] have shown that an ecological interface (EI) [3] may better support the anaesthetists' concept of decision making than a traditional trend displays. The principle of EI design is to display all information necessary for decision making in one single display: i.e. measured vital parameters, physiological background knowledge, and patient dependant context. Methods. We developed a new EI which visualises 35 relevant parameters for anaesthesia monitoring according to their function (i.e., e¡ect of anaesthesia and neuromuscular agents, temperature, £uid management, respiratory mechanics and gases, hemodynamic and oxygen supply). The latter four are assessed by intelligent alarms [4]. All parameters are generated by an anaesthesia software simulator (BODY SimulationTM ). In an experimental set-up 16 anaesthetists had to perform two simulated general anaesthetics: working 1. only with the simulator's gas and hemodynamic monitor, and 2. with the simulator's monitors in combination with the EI. During each experiment one for the subjects unknown critical incident (either a blood loss (amount: 170 ml/min, duration: 4 min) or a cu¡-leakage (leakage: 50%)) occurred at a de¢ned time and ought to be identi¢ed. To analyse the subjects' control and monitoring behaviour during the experiment eye-movements together with a think aloud protocol were recorded with a video based monocular eye-tracker system (El-Mar Vision 2000TM ). Results. The ``cu¡-leakage'' was identi¢ed in 6 of 8 cases (time (T): 215s  76s) in experiment 1 and in 7 of 8 cases (T: 65s  73s) in experiment 2. The ``blood loss'' was identi¢ed in all cases (T: 217s  72s) in experiment 1 and in 5 of 8 cases (T: 222s  189s) in experiment 2. Not identi¢ed ``blood loss''-cases when working with the EI in experiment 2 were masked through bu¡ering the e¡ect of the incident by routinely administering a high amount of lactated ringer to the ``patient'' before the incident itself starts. The eye-tracking results showed a situation dependant and interindividually very di¡erent behaviour for scanning information. Nevertheless, all trials have in common that in experiment 2 (with the EI) the new display was used as main source of information (43  19% of time). Speci¢c parameters (e.g.: expired CO2 -concentration, airway pressure, or blood pressures) could be identi¢ed as main information of interest in both experiments. Moreover, frequent ¢xations towards those display components which showed strong deviations from normal ranges indicated, that the EI especially helped in identifying an evolving critical incident.

Conclusion. The positive results for the incident ``cu¡leakage'' and the subjects' decision making behaviour when using the EI support our concept. The results have not yet shown that an EI could signi¢cantly improve the anaesthetists' decision ¢nding process, but let us assume that an attention shift to a problem with the help of an appropriate visualisation of parameter functionalities and intelligent alarms may better support the anaesthetists' decision ¢nding process. REFERENCES 1. Jungk A et al. Evaluation of three displays for hemodynamic monitoring by eye-tracking. In: Proc. of 10th Europ. Congress of Anaesthesiology, German Congress of Anaesthesiology (DAK 98), June 30 ^ July 4, 1998. Ana« sthesiologie Intensivmedizin Notfallmedizin Schmerztherapie 1998; 33: 408 2. Jungk A et al. Ergonomic Evaluation of an Ecological Interface and a Pro¢logram Display for Hemodynamic Monitoring. J Clin Monit Comput 1999 (accepted) 3. Vicente K, Rasmussen J (1992). Ecological interface design: Theoretical foundations. IEEE Trans System, Man, Cybernetics 1992; 22 (4): 589^606 4. Becker K et al. Design and validation of an intelligent patient monitoring and alarm system based on a fuzzy logic process model. Arti¢cial Intelligence Med 1997; 11: 33^53 APPROACH TO A DIALOGUE-BASED SPEECH INPUT DEVICE FOR A PATIENT MONITOR A. Jungk,1 B.Thull,1 A. Hoeft,2 G. Rau,1 1 Helmholtz-Institute for Biomedical Engineering Aachen, Germany; 2 Department of Anaesthesiology and Intensive-Care-Medicine, Bonn, Germany Introduction. The optimal way of interacting at an anaesthesia workplace has not yet been found. Precise documentation is especially required during periods of high mental workload and high manual activities (e.g. induction of anaesthesia, emergencies). Hence, documentation is often not optimally done by minutes from memory. Our hypothesis is that some of these interaction problems could be solved by speech interaction. Our ¢rst approach has been reported in [1]. Beside not acceptable low recognition rates this approach also revealed one important bottleneck of speech interaction: subjects' lost of structure in the documentation process due to lack of visual or auditive feedback. Hence, an intelligent dialogue structure is necessary to guide a user through his documentation process. Based on these experiences we designed a new speech input device for the Datex-Ohmeda AS/3TM Anaesthesia Monitor with Record-KeeperTM. Methods. As speech recognition machine the new IBM ViaVoice98TM was used. The speech input system was divided into three components: (1) a grammar of 235 words which de¢ned all possible speakable user entries, (2) a dialogue manager which analysed the semantics of spoken sentences and generated an appropriately auditory feedback, and (3) a transmission component which transmitted the entries into the monitor via simulated keyboard entries or the Datex-

ESCTAIC Abstracts 145

Ohmeda InlinkTM -Software. Beside the possibilities to correct entries, to ask for the last entry, or to cancel a dialogue, the dialogue manager allowed complete or dialogue-driven stepwise entries. E.g.: complete entry: User (U): ``0.5 Fentanyl administered,'' system response (S): ``0.5 Fentanyl administered,'' U: ``Correct,'' S: ``Finished,'' or stepwise entry: U: ``Fentanyl,'' S: ``How much Fentanyl?,'' U: ``0.5,'' S: ``0.5 Fentanyl administered,'' U: ``Correct,'' S: ``Finished.'' Moreover a user was warned if an entry did not correspond to the context of not ¢nished entries. E.g.: U: ``Fentanyl,'' S: ``How much Fentanyl?,'' U:''Weight 80,'' S: ``Weight doesn't correspond to the current context. Last entry: How much Fentanyl?'' The system was tested under optimal conditions in the lab through retrospectively documenting an anaesthetic procedure. Ten test persons (medical students, biomedical engineers) conducted each four di¡erent trials (altogether 610 entries). The test persons were trained with the specially con¢gured vocabulary. All subjects had experience with test versions of our new dialogue-based speech input device. To analyse the robustness of the system the overall recognition rate and the number of correctly ¢nished entries were calculated. The mean time to perform an experiment with the new speech input device was compared with the mean time to do it in the standard way with the keyboard. Keyboard entries were performed by one user who was familiar with the anaesthesia monitor. Results. The system recognised 2543 of 2898 words which yields an overall, word related recognition rate of 87.8%. Six hundred seven of the 610 entries were correctly ¢nished. The mean time to perform a trial with the speech input device (190s  45s) was signi¢cantly shorter than the mean time to perform a trial in a standard way with the keyboard (231s). Discussion. Compared to our ¢rst approach [1] we used a new version of the IBM ViaVoiceTM speech recognition machine, a better de¢ned grammar and semantic interpreter, and various possibilities for a dialogue-based speech input and corrections of wrong entries. The overall recognition rate increased under optimal conditions from 77.2% to 87.8% which means a reduction of the error-rate of 46.5% and which is comparable to other current systems. However, an acceptable optimal recognition rate of about 95% has not yet reached. It must be taken into account that under ORconditions the recognition rate will again decrease (52.1% in [1]). Beside these objective numbers, the evaluation has shown that the dialogue-based speech input signi¢cantly improved the usability of the system. But it could also be shown that too long system responses yield in a high workload. All of the bottlenecks mentioned in [1] could be signi¢cantly improved. However, the optimal niche for the application of speech interaction within the OR has now to be identi¢ed. Experiments in the OR will be done in a next step. REFERENCES 1. Thull et al. Experiences in designing a speech input device to a patient monitor, Proc. of the 9th Annual ESCTAIC Meeting 1998, October 8^10, 1998, Egg, Austria

TOWARDS AN ANESTHESIOLOGY INFORMATION SYSTEM (ANIS) Martin Sedlmayr, Holger Knublauch,Thomas Rose, Research Institute for Applied Knowledge Processing (FAW), Ulm, Germany Adequate and e¤cient information management is crucial in an high dependency environment such as anesthesiology. Indeed, considerable progress has been made in the development of medical information systems. However, the available solutions concentrate on special tasks, lacking the sensible integration needed for unsolicited user support and process guidance. We present here our approach towards a process-oriented departmental system for anesthesiology. ANIS is based on a model of the administrative and medical work£ow, which allows to coordinate automatically software modules for various user tasks. We consider various views on the overall process: The ``pilot's view'' of the anesthesiologist in the OR, which is covered by integrated monitoring with knowledge based decision support [3]. Another is the ``attending's view'' for support in medical management (e.g. the supervision of several ORs with the help of condensed remote monitoring). The third one is the ``administrative view'' to improve, for example, billing and statistics. The overall integration will decrease the workload, lead to a better assessment of the departments performance and can therefore improve quality and reduce cost. Our ¢rst step was to analyze the work£ow in the department and to build a domain ontology. This yielded a reference model containing a stable core (of standard processes) and a series of exceptions where the standard fails [1]. We speci¢ed the ontology and the reference model, with our recently developed framework for engineering knowledge based decision support systems [2]. The target system is a client server or distributed computing environment sharing a common data base. Each node in the network o¡ers or uses services (a service can be the data delivered by a device or the generation of a report). The user applications are software modules based on these services. We propose to use CORBA for the formation of the distributed environment and Java for the implementation of the software modules as both have shown their advantages in network programming [3]. An agent-based architecture is a suitable framework for transforming the departmental and technical models into a software architecture. The intuitive understanding of an agent is the representation of the user or patient within a virtual environment. Therefore they are also a good tool when eliciting domain expert knowledge. The actual implementation will either use true agents (i.e., agent frameworks, KQML) or an agent-oriented but otherwise conventional approach. The user interface has to follow a common look and feel to ensure the visual integration of the modules. The presented information is selected by a knowledge based algorithm depending on the current phase of the process [2]. In contrast to conventional expert systems, this selection process need only to be ``good enough'', as it does not diagnose an incident itself but displays the currently most important information. During implementation it is advantageous to separate the implementation of the user interface from the functional

146

Journal of Clinical Monitoring and Computing Vol 16 No 2 2000

software modules, because it can then be replaced more easily. A framework of graphical elements for monitoring has been developed. [3] We have built an ANIS prototype for the induction phase of anesthesia, which is based on a model of the anaesthetic processes. Currently, it serves us as a research platform for new technologies. The modular approach facilitates adding new functions and the test of new ideas. The need of information selection led to a new knowledge engineering framework which has proven its usefulness in several other projects. Also the idea of a process oriented user interface of the patient monitoring module received good remarks from anesthesiologists. We are convinced, that process orientation is a major requisite when building an integrated information system for anesthesiology. With the goal of a clinical departmental system in mind, we will expand our prototype. During this next step, we will on the one hand extend our model of the departmental processes and on the other hand focus on functions required for medical management. REFERENCES 1. Marsolek I. ``International Telecommunications Project for the Analysis and Comparison of Clinical Workprocesses in Anesthesia and Intensive Care Medicine,'' ESCTAIC 98 2. Knublauch H. ``Representing and Processing Various Knowledge Types in an Anesthesia Information System,'' ESCTAIC 98 3. Sedlmayr M. ``Integrated Monitoring with Process Adequate Visualization in a Distributed System for Decision Support in Anesthesiology,'' master's thesis (in German), University of Ulm, 1999 A NEW CONCEPT FOR SOFTWARE ARCHITECTURE OF A PATIENT DATA MANAGEMENT SYSTEM (PDMS) M. Benson,1 A. Michel,2 A. Junger,1 G. Sciuk,1 K. Marquardt,2 G. Hempelmann,1 1 Dept. of Anaesthesiology and Intensive Care Medicine, 2 Dept. of Clinical and Administrative Data Processing, Justus-Liebig-University Giessen, Germany Background. Despite the fact that the technical requirements (hard- and software) for the use of online work stations that are installed hospital wide, with a possibly complete digital patient record, have been largely ful¢lled, these systems have yet to be established for routine use. Areas where problems arise are i.e. call and provision of large amount of data of a long treatment cycle, parallel data input into the same patient record by di¡erent persons, progressive presentation of data in time with a good performance and integration of di¡erent software applications with already existing data at the hospital. With the help of the patient data management system (PDMS) of the surgical ICU ward at Justus-Liebig-University Hospital Giessen we have studied to which extent communication and message structure of the software can exert a positive in£uence on the above mentioned problems and whether our software architecture concept has been able to prove e¤ciency, suitability and acceptance. Methods. The Oracle runtime database of the PDMS is contacted via so called application servers by the ORACLE

CALL INTERFACE [2]. Every individual application server (admsrv, adtsrv, labsrv, vitalsrv, ect.) is only processing data of its corresponding section of the medical departments (program administration, patient core data, laboratory parameters, vital data etc.). The ``mapper`` as superordinate application server, coordinates the communication between the application servers and the client. Installed on the client is a ``master`` who ensures the communication on one hand with the database and on the other hand with the client applications. HL-7 [1] has been chosen for the message structure and TCP/IP as transfer protocol. Whenever possible we used standard HL-7 messages. For extensions the regulations for class enlargement of HL-7 have been considered. The surface applications were programmed with the programming languages C and C‡ using Microsoft Foundation Classes. Data is presented in a single window, data categories are shown vertically and time progress horizontally. Results. The results are supported by the experiences of the routine use of a patient data management system at a surgical ICU ward with 14 beds. The distribution of the data stream into individual application servers allows to range data on the side of the database. The installation of application servers on di¡erent physical database servers (at present 2 Pentium II/ 300 MHz-PCs), that are contacted by 19 clients via multiple threading, considerably reduces the individual response time. The standardised HL-7 message structure allows the integration of the most dissimilar data sources. This enables the application server as well as the client to both send and receive HL-7 messages via TCP/IP to/from external systems (HIS, communication server etc.). The readiness of the master to receive messages ensures an individual up to date data inventory of the client without user access or ``polling.`` A parallel handling of one patient record by di¡erent clients is ensured by a master- master communication bypassing the database. The ``master`` guarantees the integration and exchange of information of varied applications on the client (login, intensive care record, anaesthesia record, browser, ward to ward communication, medical device driver, local data bu¡er etc.). Because data are positioned as single objects in time similar to the presentation on paper, the user can fall back to his/her habitual interpretation algorithms. Parallel work on one patient record by several users is guaranteed by the communications structure any time. Discussion. The modular structure of the system into the following independent elements ``communications'' ``application'' and ``database'' makes it possible to continually improve the system and allows a variable substitution of single elements without a total reprogramming of the PDMS. The direct communication of the client via HL-7 with ``foreign'' databases and communication servers represents a totally new possibility of a variable data £ow. With this architecture every server can be regarded as client and vice versa. A real bidirectional data £ow is created, which does not require a special user access and additional polling. The described software architecture guarantees the most varied functions with good performance. In our opinion this is a constituant for a promising approach towards the development of a digital patient record.

ESCTAIC Abstracts 147

REFERENCES 1. Health Level Seven, Inc. The Standard for Electronic Data Exchange in Health Care, Version 2.2 ß 1994, 3300 Washtenaw Ave., Sweet 227, Ann Arbor, MI 48104-4250, USA 2. Oracle 7, Release 7.3.4 Dokumentation, Oracle Corporation, 500 Oracle Parkway, Redwood City, CA 94065, USA CLINICAL PRACTICE IMPROVEMENT IN CORONARY ARTERY BYPASS GRAFT SURGERY: BUILDING AN OUTCOME SPECIFIC KNOWLEDGE BASE Ulrich Bothner, MD,1;2;3 Randy Smout, MS,2;4;5 Susan D. Horn, PhD,2;4;6 1 Department of Anesthesiology and Intensive Care Medicine, University of Ulm, Germany; 2 Department of Medical Informatics, University of Utah, Salt Lake City, USA; 3 [email protected]; 4 Institute of Clinical Outcomes Research, Salt Lake City, USA; 5 [email protected]; 6 [email protected] Introduction. Coronary artery bypass graft surgery (CABG) is a frequently performed and expensive procedure with a large variation of clinical outcomes. The patient is at perioperative risk due to the life-threatening consequences of coronary heart disease, the high prevalence of comorbidity factors, and the invasiveness of the procedure. The aim of this prospective Clinical Practice Improvement (CPI) study was to examine content, timing, and impact of the individual perioperative CABG care processes. This knowledge is needed for future development of clinical protocols to achieve improved medical outcomes for the least necessary cost over the continuum of a patient's perioperative care. Methods. The CPI study comprised several steps of turning observable data into information: First, there was the application of the CPI methodology to postulate patient factors, process factors, and outcome factors. Patient factors were considered to be nonchangeable confounding or interacting variables to control for. Process factors constituted variables that can be modeled and manipulated to achieve improvements. An interdisciplinary team consisting of point-of-care physicians, nurses, and therapists agreed on a set of candidate variables. Data entry, validation, and severity of illness measurement were accomplished with the Comprehensive Severity Index (CSIÕ) software. Finally, multivariate linear and logistic regression analyses were performed to assess how patient factors and process factors predict outcomes and how associations change when severity of illness is introduced for adjustment. Results. The results of the study show that, after controlling for severity of illness, several patient factors and process factors were associated with longer length of stay on the intensive care unit and the hospital, increase in severity of illness, and postoperative complications such as death, myocardial infarction, reoperation, heart failure, atrial ¢brillation, reintubation, stroke, confusion or delirium, renal insu¤ciency and failure, as well as infections. Discussion. The results of this study provide data and knowledge for the development of decisive, executable, and dynamic clinical protocols in perioperative CABG surgery.

Protocol evaluation and determination of the clinical practice improvement changes are to be the next steps in the overall study program for better clinical outcomes and less costs. REFERENCE 1. Horn SD. Clinical Practice Improvement Methodology: Implementation and Evaluation. New York: Faulkner & Gray 1997 INFLUENCE OF DOCUMENTATION BEHAVIOR ON REPORTING OF ANAESTHESIA RELATED INCIDENTS, EVENTS, AND COMPLICATIONS Ulrich Bothner, MD,1 Bernhard Schwilk, MD,2 Department of Anaesthesiology and Intensive Care Medicine, University of Ulm, Germany; 1 [email protected]; 2 [email protected] Introduction. The German Society of Anaesthesiology and Intensive Care Medicine evaluates the standardized reporting of perioperative anaesthesia related incidents, events, and complications (IEC). As part of the data validation e¡orts, the problem of variation in documentation behavior was to be evaluated in a prospective controlled study. Methods. A study group of particularly motivated and instructed residents and attendants was scrutinized at a University teaching hospital from April through June 1996. IEC assessment was completed according to the Society's de¢nitions. Results. Out of 4242 anaesthetic procedures during the study period 839 (19.8%) were completed by the study group. In the univariate analysis, incidence of IECs was signi¢cantly (P < 0.001) higher reported by the studied physicians (36.4%) as compared to the control group (18.6%). In a multivariate analysis, while adjusting for all covariables, reported incidence of IEC was 2.4-fold signi¢cantly increased (P < 0.001). Discussion. This study is another step in validating the assessment of IECs. The results gain importance with respect

Documented incidence of the maximum IEC severity grade per anaesthetic procedure. Incidences were signi¢cantly di¡erent (P < 0.001) between control and both study groups (with or without record check in the PACU). Incidences were not signi¢cantly di¡erent (P = 0.462) between both study groups. *Total incidence of IECs may comprise several IECs with di¡erent severity grades per anaesthetic procedure.

148

Journal of Clinical Monitoring and Computing Vol 16 No 2 2000

to upcoming nationwide quality comparisons among anaesthesia providers. In addition, they give a better estimate of economic implications from IECs within the institution. CONTINUOUS THORACIC EPIDURAL ANALGESIA AND PATIENT CONTROLLED ANALGESIA FOR LATERAL THORACOTOMY. CLINICAL CONSIDERATIONS, INTRODUCTION OF THE ANALGESIC METHODS INTO CLINICAL PRACTICE AND DOCUMENTATION IN THE ACCESS 97 DATABASE PAIN MANAGER J. Hann, J. Fuchs, J. Rosenkranz, Section of Cardiac Anaesthesia, Department of Anaesthesia and Intensive Care Medicine (Director: OMR Prof. Dr.W.Ro« se), Otto-von-Guericke University Magdeburg, Medical School, Leipzigerstr.44, D-39120 Magdeburg, Germany E-mail: [email protected] Postoperative pain of more than 100 patients with lateral thoracotomies for various surgical procedures was treated either by piritramide patient controlled analgesia (PCA) or continuous thoracic epidural analgesia (CTEA) using bupivacaine or ropivacaine in combination with epidural sufentanil. Those operations are known to be very painful. Patients are anaesthetized and operated in the thoracic surgical ORs. Postoperatively they are monitored for at least one day in an ICU, from where they are transferred to various standard wards. We also included in the analgesic program some postoperative patients who only had multiple thoracical incisions for videoassisted thoracoscopic surgery (VATS) if patients wanted PCA even for minor pain levels. From literature and a one year experience using an EXCEL list we chose our documentation items for an ACCESS 97 database written by an inhouse engineer on a commercial basis. First the program opens up a patient search page. The patients are then presented in one basic patient data page and additional 9 subpages. Four pages are alternatively ¢lled out when starting or stopping (weaning) PCA or CTPA. One page is dedicated to intraoperative information. Postoperatively only 4 simple one screen pages and one larger two screen page are serially worked through. (chesttubes and urinary catheter, actual pain treatment, daily rounds, medical syringe prescription and PCA/CTEA problems). The large''daily rounds'' page is divided in items which are ¢lled in only once a day and others that are ¢lled in for every speci¢c round. There is a maximum documentation of three rounds per day. Changes in therapy type or medication should only be done after a ward round and is then documented. According to the multimodal approach we check for pain using a visual analog scale (VAS) and patient satisfaction with pain therapy (yes/no), mobilisation and gastrointestinal function. We register adverse e¡ects and laboraratory ¢ndings like serum creatinine and serum urea to supervise renal response to additional NSAIDS like diclofenac or metamizole. In PCA we document opiate use in mg between the rounds, calculate mean hourly demand and check for the ratio of total demands to positive demands of opiate boluses to evaluate our PCA mode settings. In CTEA the spread of temperature or pinprick sensation loss is documented. We started to do local anaesthetic serum levels once a day to prevent side e¡ects of prolonged treatment.

There is a sensitive balance between documentation needs to describe a patient clinically for a 4 to 5 day treatment phase and the amount of time and complexity demanded from the examining anaesthesiologist. The accumulated data is stored centrally in an ACCESS 97 server database. The data acquisition is done on ward PCs or documented on a paper checklist to be typed in back in our clinic if there is no accessible PC available. All our wards have login controlled network access that can be used with a laptop. There are four standard reports that can be printed locally or on remote network printers. The patient checklist enables to see and print basic patient data and the results of the last round. The patient pro¢le gives the results of all the patients over time to enable decision making and to control the completeness of data. We feel that for all patients a ¢nal report should be written to be added to the patients record. A prepared WORD 97 Macro split in PCA and CTEA patients is ¢lled with structured data to speed up report making without obligatory use of secretary. The prescription shows the used mixture of medication. For scienti¢c data analysis and plotting ACCESS 97 o¡ers even SQL untrained users rather simple tools to create tables which can be exported to statistic and plotting programms. PAIN MANAGER is learnt intuitively with user training only necessary on a few real patients. We started developping the database in Jan 99 and hope to meet our goal to ¢nish fund raising, tests and ¢nal revision to have a commercially distributable German and English version by Dec 99. MEASUREMENT OF ISOFLURANE IN OXYGEN: COMPARISON OF THE CAPNOMAC ULTIMA AND AN INSTRUMENT BASED ON A FLUIDIC BISTABLE AMPLIFIER V. Anderson, V. Kulkarni, D. G. Ross, Department of Anaesthesia & Intensive Care, Grampian University Hospitals, Aberdeen, Scotland Introduction. Anaesthetic vapour concentration can be measured using a number of techniques including Infra-Red spectroscopy, Photo-acoustic devices, Raman spectrometry and Mass spectrometry. A binary mixture of oxygen and an anaesthetic agent like iso£urane or halothane can also be measured using a £uidic bistable ampli¢er [1]. We wished to compare measurements of iso£urane in oxygen using a Capnomac Ultima (Datex) [2] with those of ARIGA (Aberdeen Royal In¢rmary Gas Analyser) ^ an instrument based on a £uidic bistable ampli¢er. Methods. The anaesthetic circuit consisted of a Manley Pulmovent. The ventilator was in manual mode so supplying a constant gas stream. Sampling was performed from two ports at the junction of the inspiratory and expiratory limbs of the circuit. Following calibration, the Capnomac Ultima was connected to one sample port, while the other sample port was linked to the pneumatic module of ARIGA. The electronic module of ARIGA counts and displays the frequency of oscillation of the £uidic bistable. A set of ¢fteen measurements was made with both instruments at a number of vapouriser settings. Results. The mean of each set of ¢fteen readings from both instruments was calculated. The Figure shows the mean

ESCTAIC Abstracts 149

frequency of oscillation of ARIGA plotted against the Capnomac Ultima readings.. There was a good correlation between the two instruments with a correlation coe¤cient of 0.989. Conclusions. The results of this study indicate that ARIGA, an instrument based on the £uidic bistable ampli¢er, compares favourably with the Capnomac Ultima when measuring iso£urane in a binary mixture with oxygen. This study was funded by the Aberdeen Royal Hospitals Trust Endowment Fund. REFERENCES 1. Rodgers RC, Ross DG. Fluidic bistable ampli¢er as a vapour concentration sensor.Proceedings, Anaesthetic Research Society. Br J Anaesth 1991; 66: 401^402 2. McPeak H, Palayiwa E, Madgewick R, Sykes MK. Evaluation of a multigas anaesthetic monitor: The Datex Capnomac. Anaesthesia 1988; 43: 1035^1041 SIXTEEN YEARS EXPERIENCE WITH AUTOMATIC DOCUMENTATION IN THE OR A. F. de Geus, PhD, G. E.Wiersma, R. Huet, MD, Dept. of Anesthesiology, University Hospital Groningen,The Netherlands Introduction. In 1983 the ¢rst Carola system, an Automated Record Keeper (ARK) for cardiac surgery, was introduced in the OR. Currently four systems document 1900 operations a year, and links exist with a commonly used blood gas analysis device, the laboratory, the database of cardiology and the hospital information system. All data are stored automatically in a database, which is accessible by an intranet browser. In its long history Carola maturated from a dedicated microprocessor based ARK with a plotter as output device, to an open network based Anesthesia Data Management System (ADS) running on a Unix server, and employing a 20 inch colour display. Doing it yourself. Carola is an own development, without any support from industry. It is often questioned whether hospitals should develop their own ADS, or that they should leave it to the industry. What are the advantages and the disadvantages? What investments in equipment and human resources are necessary?

What principal design decisions have to be taken? Can the quality be as good as that of industrial products? When the initial system has been developed, what about maintenance and dependency on the people that develop and support the system? How to cope with the fast evolving computer industry? The Carola project. The Carola project has proven that it is feasible to design, develop and maintain your own ADS. With the fast evolution of computer technology the complexity of the system has increased enormously. Though started as a stand alone system in the OR, the number of connections with external systems and devices outside the OR is still growing. Carola is completely embedded in the general hospital automation structures. All internal communication is also network based: the graphical displays are X-terminals; equipment, printers and the Carola database are all on the network. The major issue was and still is to meet changing local requirements. When equipment in the OR is replaced or the layout of the OR is changed, the integration with Carola is one of the major decision criteria. Maintenance. The obvious advantage of a network based system is £exibility and accessibility. The drawback is the dependency on the computer network. This turns out not to be a problem, there is little downtime. The systems are up 24 hours a day. Network failure occurs only once or twice a year. Because all network infrastructure is continuously monitored, problems are reported immediately and take less than 15 minutes to be solved. The hardware used for Carola is very reliable. The four HP Unix workstations are now 24 hour operational for over 5 years. Only once a disk crash occurred. The operating system HP-UX never crashed! The X-terminals had no problems either, only once a boot did not succeed. The Carola software is mostly stable. In normal conditions once or twice a month assistance is needed from the development team. Only when complete or partial reworks are done on the software, there is a short period of 1-2 weeks, in which in spite of all the testing, bugs need to be removed. This never endangered the basic functionality of Carola. The frequency of these major updates is once a year. All operations are written automatically to a database. It occurs unattendantly and only requires some attention on software updates. The connections to external systems do require maintenance. Functionality and protocols are sometimes changed without notice. It is vital to check the functionality of these connections on a regular basis. Discussion. What makes an ADS a success? There are many factors of in£uence. An experiment to expand Carola for use in general surgery was less successful and aborted. It has given much insight what the minimal requirements are to develop an ADS, and to introduce it in the OR. A major factor is the necessary time investment and cooperation with users when the system is introduced en tested in daily practice. Manual documentation and automatic data management di¡er and require a di¡erent attitude in handling patient data. Another factor is that partial solutions do not work. The Carola clone was installed in only 3 of the remaining 20

150

Journal of Clinical Monitoring and Computing Vol 16 No 2 2000

OR's. This reduces the necessary experience among the users, essential for motivation and feedback. When problems are not known or when there is a shortage of manpower the development gets stuck, and acceptability diminishes. We believe, however, that above observations are not restricted to an own developed system, but that they apply equally well to industrial ones. AUTOMATED PEROPERATIVE BLOOD DONOR PRODUCT REGISTRATION R. Huet, A. F. de Geus, G.W.Wiersma, Dept. of Anesthesiology, University Hospital Groningen,The Netherlands Introduction. Blood product ordering and administering encompasses many steps liable to error, especially in emergency situations when many diverse products have to be accounted for. Surprisingly, the registration of these products depends world wide on failable human control and security checks. It seems challenging to automate the blood product transfusion procedure. Therefore, the usuability and pitfalls of an automated registration system for blood product transfusion was exploited. Methods. The £owchart of the peroperative blood donor product ordering, reservation and administering procedure was translated into an automation process and evaluated in routine practice within the setting of cardiac surgery. The procedure was divided in several steps: a) ordering speci¢c blood donor products, b) patient speci¢c labeling of blood product type, number, and blood group, c) patient speci¢c identi¢cation of available blood products next to the patient, and d) labeling of the responsible physician, e) documentation of administered of blood product, f) feedback to transfusion lab of administered products, and g) documentation of administered products at transfusion lab. Blood donor products are barcoded internationally for product type, blood group and unit number. Upon reservation a patient ID label is attached on the unit (no barcode). To enter data of blood donor products, a barcode reader coupled to the Anesthesia Information System (AIS) is essential. We use a hospital developed AIS (CAROLA) operational in daily practice since 15 years on 4 cardiac theatres. Since 5 years, Carola runs on a UNIX operating system and employs a 2000 color display X-terminal. At the start of the anesthesia procedure, the hospital identi¢cation code of the patient is manually entered into the AIS. Thereafter, the AIS interrogates the hospital information system for initially reserved blood products. Subsequently, the AIS informs the anesthesia team of reserved blood products. If, during the operative course a blood product is necessary, the barcodes of the blood group, the unit number and the product type are identi¢ed with the barcode reader. The barcodes are checked and an auditory signal con¢rms the identi¢cation and veri¢cation process. If group, type and number is correctly identi¢ed, the anesthesia team can enter the infusion velocity, the total quantity or both. At the end of the operation all given blood product information, the responsible physician, and the time of administering is available for documentation at the blood transfusion lab. Results. The procedure is now operational, and some problems in the automation of the procedure have been

identi¢ed. The level of automation at the transfusion lab and the hospital information system attenuated a full exploitation of the possibilities. The identi¢cation and veri¢cation process of the blood product in the OR was time saving, especially in babies and emergencies. In average about 10 seconds were necessary to scan and register the products. This reduced the usual procedure whereby 2 anesthesia team members double check the blood products identi¢cation sheets and enter the 7 digit unit numbercode in the computer. Furthermore, occasional fustration with mislaid or disappeared identi¢cation sheets during emergencies was prevented. Discussion. Automated blood donor product ordering and documentation contributes to the safety of blood product registration by reducing human error. This could not be proven within this context. Unfortunately, ordering of additonnal blood products could not be automated yet. A requisite for the automated handling of blood donor products is an automated blood transfusion lab and on line information and ordering via the hospital information system. Additionnally, in our institute, the provided documentation of actual administering could not yet be automatically processed by the blood transfusion laboratory. In the OR the automation reduced the workload of the anesthesia team, especially during emergency circumstances, and in babies. In conclusion, automation of peroperative blood donor products is feasible, saves time and eases work in the OR. But automation of peroperative blood product registration needs considerably software and hardware tuning of several departments to be e¤cacious. A COMPLEX-SYSTEMS APPROACH TO SAFER DRUG ADMINISTRATION IN ANAESTHESIA Craig Webster, Alan Merry, Louise Larsson, Ken McGrath, Department of Anaesthesia, Green Lane Hospital, Auckland, New Zealand Introduction. Iatrogenic harm is indisputably a cause for concern in all ¢elds of medicine [1] and anaesthesia is no exception [2]. At Green Lane Hospital we have developed a new Injectable Drug Administration System (IDAS) which incorporates salient cueing and tracking features speci¢cally designed to reduce error and facilitate performance. The IDAS is based on considerations of cognitive ergonomics, principles of systems safety developed in the nuclear and aviation industries, and the lessons of successful error reduction strategies currently used in other areas of modern anaesthesia. Methods. As part of the clinical assessment of the IDAS we are conducting an incident monitoring study which captures data from both IDAS and conventional methods of drugs administration. This study identi¢es both a reliable numerator and denominator of the incidence of error by asking anaesthetists to complete an incident form for every anaesthetic, the vast majority of which identify ``no incident'' as having occurred. In addition, we speci¢cally identify precursors to error (or pre-errors) as an important and conventionally overlooked source of information on the aetiology of error [3]. Results. In the ¢rst 18 months of the incident study 6156 incident forms have been returned from 7654 anaesthetics ^ a response of over 80% (comprising 7286 conventional and 368

ESCTAIC Abstracts 151

IDAS vs. conventional methods

Errors Pre-errorsa Total anaesthetics n a

Conv. (% of n)

IDAS (% of n)

42 (0.6%) 27 (0.4%)

1 (0.3%) 6 (1.6%)

7286

368

Fisher, p < 0.005.

IDAS anaesthetics). During this time 57 errors have been reported ^ 42 of which related to an intravenous drug. Six errors relating to a gas or vapour and 8 infusion errors also occurred and were excluded from further analysis as the IDAS currently has no facility for these modes of administration. Intravenous errors occurring with the conventional system comprised 14 administrations of an incorrect dose, 9 substitutions of one drug for another, and 5 cases of omission of a drug. Factors reported as contributing to these errors included 10 errors related to similar or indistinct labelling, 7 related to problems of communication, 6 related to distraction, and 5 related to haste or fatigue. The single IDAS error involved repetition of an administration when 4 core IDAS operating procedures were ignored and communication broke down between the two anaesthetists involved when one was absent. Signi¢cantly more pre-errors occurred with the IDAS than with conventional methods (table 1). Incident forms speci¢cally identi¢ed 4 of the 6 IDAS pre-errors as cases where IDAS safety features stopped an error from occurring. Discussion. Our data suggest that anaesthetic error occurs surprisingly often. The IDAS represents one of the ¢rst attempts to apply a complex-systems approach to anaesthetics. The greater number of pre-errors seen with the IDAS is consistent with the complex-systems principle of detection and prevention of latent system problems before they manifest as errors, thereby trapping problems at the pre-error level. In the long term we expect this feature to show a de¢nitive safety advantage for the IDAS. REFERENCES 1. Leape LL, Brennan TA et al. The nature of adverse events in hospitalized patients. N Engl J Med 1991; 324: 377^384 2. Merry AF, Peck DJ. Anaesthetists, errors in drug administration and the law. NZ Med J 1995; 108: 185^187 3. Reason J. The contribution of latent human failures to the breakdown of complex systems. Phil Trans R Soc Lond B 1990; 327: 475^484 OPTIMISING THE INTERVAL BETWEEN NON-INVASIVE BLOOD PRESSURE DETERMINATIONS Ross Kennedy, Department of Anaesthesia, Christchurch School of Medicine, Christchurch, New Zealand Background. The frequency of automated non-invasive blood pressure measurements is a balance between potentially deleterious e¡ects of rapid cycling and the potential delay in detecting changes with a long cycle time. This study aimed to explore the e¡ect of increasing cycle times and to determine if

simple rules based on pulse rate changes could determine when extra measurements should be made. Methods. A computer model which generated ``blood pressure'' changes was used. Blood pressure measurements were made at intervals between 1 and 10 minutes. These measurements were analysed by a simple trend detection algorithm (Triggs Tracking variable, TTV) which indicated when a signi¢cant change had occurred. Using data from 100 cases lasting more than 60min in which non-invasive blood pressure measurement had been used, pulse rate changes (TTV >0.6 or 0.9) were used to attempt to predict a signi¢cant blood pressure change as de¢ned by a change of 10 mmHg, 20 mmHg or a ``signi¢cant'' trend in blood pressure. Results. In the computer study the detection time increased linearly as the interval between determinations increased. The average time to detect a change was 3.7 min with a 1 min cycle, 3.9 min at 2 min and 6.1 min with a 5 min cycle. The lack of a pulse rate change is a reasonable predictor of the lack of a blood pressure change (negative predictive value 0.9 for a 20 mmHg change and 0.7 for a 10 mmHg change). Similarly a high TTV value for pulse rate predicted a high TTV value for blood pressure (positive predictive value 0.7^ 0.8). Pulse rate changes were poor predictors of blood pressure changes (positive predictive value 0.11^0.36). Conclusions. Increasing the interval between blood pressure determinations increases the delay in detection of a change but this increase in detection is about half the increase in the interval. Provided abrupt changes in blood pressure are not anticipated, a one or two minute cycle time for o¡ers little advantage over a ¢ve minute cycle but may increase the risk of complications. Extra measurements made on the basis of pulse rate changes will detect a high proportion of blood pressure changes, however many of these extra measurements will not detect any change. REFERENCE 1. Kennedy RR. A modi¢ed ``Triggs Tracking Variable'' as an advisory alarm during anaesthesia. Int J Clin Monit Comput 1995; 12 (4): 197^204 CLOSED-LOOP CONTROL OF ANAESTHESIA USING THE BISPECTRAL INDEX A. R. Absalom, G. N. C. Kenny, University Department of Anaesthesia, Glasgow Royal In¢rmary, Glasgow, Scotland We have developed a system for closed-loop control of anaesthesia using the bispectral index (BIS) as the input signal. We present the results of a study into its performance for controlling maintenance of anaesthesia during minor surgery. Background. The system consists of an A-1000 BIS monitor (Aspect Medical Systems, software v. 3.3), a personal computer, and a target-controlled infusion (TCI) system. The BIS is downloaded every 5 seconds and passed to a proportionalintegral control algorithm, which calculates adjustments to be made to the blood propofol concentration. The TCI system administers the required infusion rates, thus closing the loop. Initially the algorithm used the BIS and blood concentration to calculate adjustments to the target propofol

152

Journal of Clinical Monitoring and Computing Vol 16 No 2 2000

concentration, but after 10 cases it was altered to also use brain propofol concentration, to allow e¡ect-site ``steering.'' Methods. After ethics committee approval was granted, twenty healthy adults who gave written informed consent were enrolled. All were scheduled to undergo day-case body surface surgery. Prior to induction of general anaesthesia we began recording the BIS. General anaesthesia was induced using the TCI system under the control of the anaesthetist. During induction a TCI infusion of remifentanil was commenced (target concentration 4 ng/ml). A laryngeal mask airway was inserted and mechanical ventilation with oxygen and air commenced. After skin incision, when the anaesthetist was satis¢ed that anaesthesia was adequate, automatic control was initiated. The BIS at that moment was used as the target BIS value for the closed loop system. BIS, blood and brain propofol concentrations were recorded automatically. Physiological variables and presence of any movement were recorded manually. At the end of surgery the infusions were stopped, but all variables were recorded until consciousness was regained. System performance was assessed using standard measures [1]. Patients completed a satisfaction questionnaire later in the day. Results. Mean duration of automatic control was 30 (range 12^88) minutes per patient. During this time the actual BIS was within 20% of the target value 92% of the time, and cardiovascular and respiratory variables were stable. One patient moved during anaesthesia when the remifentanil infusion was stopped inadvertently. Median Performance Error was 0.00%, Median Absolute Performance Error (MDAPE) was 5.88%, wobble was 5.64% and mean o¡set was 0.01%. MDAPE and wobble were 20% better with the new algorithm than with the old algorithm. All patients were satis¢ed with their anaesthetic, and there was no occurrence of awareness. Discussion. Mortier has described a system using the BIS for automatic control of sedation [2]. We have produced a system for automatic control of anaesthesia that has provided stable physiological conditions in a group of patients undergoing general anaesthesia for minor surgery. The accuracy of control of the BIS for general anaesthesia compared favourably with that reported by Mortier for propofol sedation. REFERENCES 1. Varvel JR, Donoho DL, Shafer SL. Measuring the predictive performance of computer-controlled infusion pumps. J Pharmacokin Biopharm 1992; 20: 63^94 2. Mortier E, Struys M, De Smet T, Versichelen L, Rolly G. Closed-loop controlled administration of propofol using bispectral analysis. Anaesthesia 1998; 53: 749^754 EEG VARIABLES PREDICT RESPONSES CAUSED BY STIMULI DURING PROPOFOL SEDATION Matsuyuki Doi, MD, Shigehito Sato, MD, Gavin N. C. Kenny, FRCA, 1 Department of Anesthesiology and Intensive Care, Hamamatsu University School of Medicine, Japan; 2 University Department of Anaesthesia, Royal In¢rmary, Glasgow, Scotland The objective of the present study was to compare the e¤cacy of the auditory evoked potential index (AEPindex), the

Pk values of the four variables.

AEP BIS SEF MF a

Verbal/ eye open

Forehead/ eye open

Suction/ Suction/ movement cough

0.85a 0.73a 0.65 0.60

0.83a 0.74a 0.63 0.65

0.79a 0.75a 0.70 0.61

0.73a 0.68 0.65 0.63

p < 0.01 larger than 0.5.

Bispectral index (BIS), 95% spectral edge frequency (SEF) and median frequency (MF) in predicting responses caused by various stimuli during propofol sedation in intensive care patients. Methods. After obtaining approval from the Ethics Committee and informed consent, we studied 20 patients undergoing propofol sedation. Using a TCI system, target blood propofol concentrations were set at 0.5, 1.0, 1.5 and 2.0 mg/ml at random. After at least 15 min had elapsed after achieving the target concentration, patients were given the following stimuli; verbal command to open eyes, light tap on the forehead, and tracheal suctioning. Opening eyes in response to verbal command or forehead tap were considered positive responses. Cough, massive movement, 10% increase in heart rate or in systolic blood pressure were considered tracheal suctioning induced responses. The BIS, MF and SEF were measured using an EEG monitor (A-1000, Aspect Medical Systems). The AEPindex was obtained using a similar system to that described in our previous study [1]. Each variable was recorded simultaneously and values averaged over 15 sec intervals. Ability of each variable at 60 sec before stimuli to predict the responses was evaluated using prediction probability (Pk). A Pk value was considered to be larger than 0.5 when p < 0.01. Results. With regard to four responses; eye opening against verbal command and forehead tap, cough and movement against tracheal suctioning, the AEPindex provided the largest Pk values (0.85^0.73) and the BIS gave the second largest values (0.75^0.68). Neither SEF nor MF provided Pk values larger than 0.5 for the four responses. No variable could predict an increase in heart rate or systolic blood pressure. We conclude that, of the four variables, the AEPindex was the most reliable measurement for predicting responses caused by the stimuli during propofol sedation. REFERENCE 1. Br J Anaesth 1997; 78: 180^184

ESCTAIC Abstracts 153

THE PERFORMANCE OF EEG BISPECTRAL INDEX AND AUDITORY EVOKED POTENTIAL INDEX TO PREDICT LOSS OF CONSCIOUSNESS ^ A STATISTICAL COMPARISON S. Schraag, U. Bothner, R. Gajraj, G. N. C. Kenny, M. Georgie¡, Dept. of Anaesthesiology, University of Ulm, D-89070 Ulm, Germany and Dept. of Anaesthesia, University of Glasgow, Royal In¢rmary, Glasgow, UK Introduction. The bispectral index (BIS) of the electroencephalogram (EEG) and middle latency auditory evoked potentials (AEP) are likely candidates to measure the level of unconsciousness and thus may improve outcome with respect to the incident of possible awareness during anaesthesia [1][2]. We statistically investigated the predictive performance of both measures obtained simultaneously to distinguish between the conscious and unconscious state during Propofol infusion. Methods. Twelve patients undergoing lower limb orthopaedic surgery during regional anaesthesia additionally received propofol by target controlled infusion (TCI) for sedation. The EEG bispectral index (BIS) and the auditory evoked potential index (AEPI), a mathematical derivative of the morphology of the AEP waveform, were recorded simultaneously in all patients during repeated transitions from consciousness to unconsciousness. Logistic regression procedures and receiver operator characteristic (ROC) analysis were used to estimate the probability of correct prediction and calculate sensitivity and speci¢city, respectively. Non-linear association of the state of consciousness with BIS and AEP values was calculated according to the following equation: Pˆ

1 1 ‡ emeÿ… 0 ‡ 1 x1 †

‰3Š:

Results. In the logistic regression model, both BIS and AEPI were signi¢cant predictors of unconsciousness (p < 0.0001), whereas the area under the ROC curve for AEPI (0.968) was larger than for BIS (0.922), indicating a better discriminatory performance (Figure). This ¢nding is supported by the ob-

Probability value

AEPI

BIS

p5% p50% p95% Pk Sensitivity Speci¢city

56.7 48.1 39.4 0.970 85.71% 93.06%

91.9 77.1 62.3 0.922 83.10 86.11

tained results for sensitivity and speci¢city. Probability values for a 50% (5%, 95%) correct prediction of unconsciousness are listed in the Table. Discussion. We conclude that both BIS and AEPI are suitable means for monitoring the level of unconsciousness during propofol infusion. However, AEPI proved to o¡er more discriminatory power in the individual patient and therefore may better describe the threshold situation of the transition phase. REFERENCES 1. Kearse LA et al. Anesthesiology 1998; 88: 25^34 2. Davies FW et al. Anaesthesia 1996; 51: 107^113 3. Walker SH, Duncan DB. Biometrika 1967; 54: 167^179 THE EFFECT OF FENTANYL BOLUS DURING ISOFLURANE OR PROPOFOL ANAESTHESIA ON THE RAT ELECTROENCEPHALOGRAM (EEG), AUDITORY EVOKED POTENTIAL (AEP), AND THE PEDAL WITHDRAWAL RESPONSE (PWR) L. Antunes,1 J. Roughan, P. Flecknell, Comparative Biology Center, Medical School, Framlington Place, Newcastle uponTyne; 1 Sponsored by PRAXIS XXI program, Science and Technology Foundation, Lisbon In laboratory animals as in humans the correct assessment of the depth of anaesthesia remain a challenge. The processed EEG and the AEP have been studied to assess the depth of anaesthesia in man and animals. In humans, the primary auditory cortical waves of the AEP re£ect the degree of hypnosis (Thornton, 1991), but the value of electroencephalographic parameters are still not clear. Fentanyl is commonly used as an analgesic agent in animals and humans. When additional analgesia is provided during anaesthesia by fentanyl supplementation, di¡erent e¡ects on the AEP have been reported including no dose-dependent e¡ects (Schwender et al., 1993) and lack of stimulation of the AEP responses by surgical stimulation (Schwender et al., 1994). There appear to be no reports of the e¡ects of fentanyl on EEG and AEP parameters in rats. For both legal and ethical reasons, the level of care during anaesthesia of experimental animals should reach the same standards has those used in human anaesthesia. We have studied in a preliminary study the e¡ect of intravenous fentanyl (10^20 mg/kg/injection) on several parameters of the rat AEP (latency of waves at P10, N18, P40 and N60) and EEG (median frequency, spectral edge frequency, total power and burst suppression rate), during a stable anaesthetic regime with iso£urane (0.7 MAC, 1% concentration), and propofol (20^40 mg/kg/hour).

154

Journal of Clinical Monitoring and Computing Vol 16 No 2 2000

Ten male Wistar rats 429g (SD  26 g) were anaesthetized with iso£urane (5% induction) and silver ball chloride electrodes were implanted. The lateral tail vein and femoral artery were canulated and the trachea intubated. Body temperature, heart rate, blood pressure, respiratory rate and capnographic data were recorded. The pedal withdrawal re£ex (PWR) was assessed as a clinical sign of depth of anaesthesia. A Wilcoxon test was employed (p < 0.05) to compare the various parameters during periods of positive and negative PWR. After a bolus of fentanyl during iso£urane anaesthesia, N20 and P40 latency increased, from 20.4 milliseconds (SD  3.5) to 21.5 (SD  1.5) and from 37.9 (SD  3.6) to 41.8 (SD  1.3) respectively, p < 0.05. Under propofol anaesthesia P40 and N60 increased signi¢cantly from 40 (SD +/- 2.9) to 43.9 (SD +/- 2.1) and from 70 (SD +/- 10.9) to 90 (SD +/- 15.7), respectively. Our study demonstrated changes in the middle latency AEP after a bolus administration of fentanyl to abolish the PWR. The studied EEG parameters, under fentanyl supplementation in combination with propofol or iso£urane anaesthesia, do not express any relation with the classical concept of depth of anaesthesia, movement or lack of movement. REFERENCE 1. Schwender D, Golling W, Klasing S, Faber-Zullig E, Poppel E, Peter K. E¡ects of surgical stimulation on midlatency auditory evoked potentials during general anaesthesia with propofol/fentanyl, iso£urane/fentanyl and £unitrazepam/fentanyl. Anaesthesia 1994; 49 (7): 572^578 2. Schwender D, Rimkus T, Haessler R, Klasing S, Poppel E, Peter K. E¡ects of increasing doses of alfentanil, fentanyl and morphine on mid-latency auditory evoked potentials. Br J Anaesth 1993; 71 (5): 622^628 3. Thornton C. Evoked potentials in anaesthesia. Eur J Anaesthesiol 1991; 8 (2): 89^107

KNOWLEDGE-BASED DECISION SUPPORT IN AN ANAESTHESIA INFORMATION SYSTEM Holger Knublauch, Martin Sedlmayr,Thomas Rose, Research Institute for Applied Knowledge Processing (FAW), Ulm, Germany Introduction. As clinical monitoring technology advances, the amount of data available to the medical personnel increases. Although this data can contribute to an improved assessment of the patient's state, it is likely to cause data overload, which strains the anaesthesiologist. In the research project ANIS, we build a computer system to reduce this data overload. Methods. The main philosophy behind ANIS is that clinical decision making is based on a small ^ but intuitively selected ^ part of the overall available information only. Aiming at decision-support for anaesthesia, ANIS provides an integrated monitoring display which draws the personnel's attention onto those information, which are most important in a given situation. For that purpose, the system collects patient data (monitoring, CPR), evaluates them within the current context and then semi-automatically selects appropriate graphical or textual displays. Since the level of intelligence such a computer system can expose is determined by the quality of the underlying knowledge and its elicitation strategy, we paid attention to a traceable and e¤cient knowledge engineering process. Results.We have built an ANIS prototype for the induction phase of anaesthesia, which is based on a model of the anaesthetic processes. For each of the identi¢ed process phases ^ information, monitoring, medication, and ventilation ^ we have implemented intuitive graphical components and screen layouts. Sample screens for monitoring and drug planning are shown below. Using touch screen buttons or voice control, clinicians can switch between these screens and access detailed information, such as the patient's history and pharmacokinetic simulations. If the system detects a change in the patient's state, which might indicate an evolving critical situation, the key factors are emphasized and presented to the

Sample Screens from the ANIS prototype (monitoring phase, drug planning and the knowledge editor).

ESCTAIC Abstracts 155

human decision-maker. Furthermore, ANIS is embedded into the clinical intranet, so that, e.g., the state of the current operations can be centrally supervised by the resident or information from other departments can be included. As a secondary result of the project, we have developed a framework for engineering knowledge-based decision support systems. Starting with a set of informal knowledge sources (e.g. text books), domain experts and knowledge engineers collaborate in the de¢nition of a set of concepts and relationships, which are a compromise between the domain expert's intuitive understanding and a su¤ciently formal syntax, which can be used by inference algorithms. Based on such a knowledge model, speci¢c information on the validation of patient data, the assessment of ¢ndings, the recognition of emerging incidents, e¡ective drug doses and standard treatment plans can be entered by means of comfortable knowledge editors (right part of ¢gure). Using domain terminology, the knowledge base can be easily evaluated with simulated or recorded patient data and scenarios. Acknowledgements. We would like to thank C. Backhaus, B. ClaÞen and W. Friesdorf for collaborating in the ANIS project. The project is being supported by the German Ministry of Research and Science. CARDIAC PERFORMANCE MONITORING IN THE PATIENTS WITH A LEFT VENTRICLE ASSIST DEVICE (VAD) A. Eremenko, V. Kislukhin, N. Krivitsky, Russian Research Center of Surgery, Moscow Recently valve-free centrifugal pumps or arti¢cial heart ventricles have been used in patients with severe cardiac failure refractory to inotropic drug therapy. Usually, recovery of the myocardium and heart pump function requires a certain period of time and dynamic assessment of integral and separate performance of the heart and VAD is of great importance. The aim of the study was to develop and apply an ultrasound velocity dilution technique to cardiac and VAD performance assessment in the patients with severe left ventricular failure in the early postoperative period after cardiopulmonary bypass surgery. Materials and methods. Integral performance of the heart and VAD was measured by ultrasound dilution method using HD-01 (HT-109) ultrasound £owmeter ( Transonic Systems, USA) with clamp on probe placed on arterial line of VAD. The probe registers blood £ow through the tube and changes of relative sound velocity of the blood after two i.v. injections of 20 ml saline solution and 5 ml hypertonic saline (10% NaCl). The indicator dilution curves were used for total output (heart + device) calculation using the formula: TCO = (k1 ÿ k2)/(S1/V1 ÿ S2/V2), were k1 and k2 ^ calibration factors, S1 and S2 ^ areas under the indicator dilution curves and V1 and V2 ^ volumes of injections. Patient's cardiac output can be obtained with the formula: CO = TCO ÿ VAD£ow, and the letter is measured by ultrasound £owmeter. Results. The approach described above was used in two patients during left ventricular bypass (Biopamp, Medtronics,USA). Comparison of total cardiac output measured by

thermal and ultrasound dilution methods showed good correlation (r = 0.96, n = 21) and absolute di¡erence between CO values was less than 10%. Conclusions. Ultrasound velocity dilution method can be used as noninvasive alternative to thermodilution method in patients with left heart assist device. It gives possibility of separate assessment of patient's heart and ventricular assist device performance. SIMULATED DIFFICULT INTUBATION: COMPARISON OF A NEW VIDEOOPTICAL INTUBATION STYLET VS. THE CONVENTIONAL STYLET IN A MANNEQUIN Peter Biro,1 MarkusWeiss,2 Andreas Gerber,2 Thomas Pasch,1 1 Department of Anaesthesia, University Hospital Zurich, Switzerland; 2 Department of Anaesthesia, University Children's Hospital Zurich, Switzerland Introduction. Di¤culties with tracheal intubation are usually caused by impaired direct visualisation of the vocal cords. Particularly, unanticipated di¤cult intubation can lead to critical situations in patients who are at risk for hypoxaemia, gastric regurgitation, or who have limited cardio-pulmonary capacity. Usually, in such situations a conventional intubation stylet (CIS) or a gum-elastic bougie is used for blind probing of the glottic entrance. The video-optical intubation stylet (VOIS) is a new device [1], which transmits the view from the stylet tip directly onto a bedside video-display and facilitates the guidance of the endotracheal tube (ETT). The purpose of this study was to assess the handling, reliability and e¤cacy of VOIS in a standardised di¤cult tracheal intubation setting and to compare it to the application of CIS. Methods. Forty-¢ve anaesthetists (15 consultants, 15 residents, 15 anaesthesia nurses) were asked to attempt tracheal intubation using the CIS and the VOIS after 1 demonstration by the examinator and 1 practical attempt by the subject in order to familiarise with the technique. Laryngoscopy was performed by the examinator with a Macintosh laryngoscope blade, who created a grade 3 view (according the classi¢cation by Cormack and Lehane) [2] in a standardised fashion. Then each subject performed both techniques 5 times in succession in a randomised order. For each attempt, time to place the ETT was documented. Failed intubation was de¢ned as an intubation duration exceeding 60 sec or a resulting oesophageal or endobronchial ETT position. Finally, the subjects were asked to express the degree of di¤culty for each technique according a verbal rating scale from 1 (very simple) to 5 (extremely di¤cult). The results were statistically evaluated using the Mann^Whitney U-Test and the Chi2 -Test with a probability level of p < 0.05. Results. Mean intubation duration for VOIS was 20.4  7.7 sec, and for CIS 10.2  3.3 sec (p < 0.01). With VOIS, the trachea was correctly intubated in all 225 attempts. In the CIS group 44 (30.4%) oesophageal as well as 44 endobronchial intubations occurred (p < 0.01). Estimated degree of di¤culty was signi¢cantly lower for VOIS (2.1  0.6) compared with CIS (2.9  1.0) (p < 0.01). There was no signi¢cant association between the success rate and intubation duration with the professional degree, years of experience, or the order of the intubation technique.

156

Journal of Clinical Monitoring and Computing Vol 16 No 2 2000

Conclusions. Although mean intubation duration with VOIS was twice as long as with CIS, the total intubation time was in an acceptable range. The VOIS enabled the anaesthetist to recognise an inappropriate ETT position and to improve it immediately. Besides this, the video-optical intubation technique showed a considerably lower degree of technical di¤culty than the conventional method. Due to the e¤cient avoidance of oesophageal intubations and endobronchial ETT malpositions, the video-optical intubation stylet may become a reliable and e¡ective tool for the management of the unexpected di¤cult intubation. REFERENCES 1. Weiss M. Video-intuboscopy: A new aid to routine and di¤cult tracheal intubation. Br J Anaesth 1998; 80: 525^527 2. Cormack RS, Lehane J. Di¤cult tracheal intubation in obstetrics. Anaesthesia 1984; 39: 1105^1111 DEVELOPMENT OF AN INTRAOPERATIVE CT-SCAN BY USING A QUICK-MOTION SCENARIO ^ RESULTS OF A STUDY K.-U. Neth, W. Friesdorf, Department for Human Factors Engineering and Ergonomics 1999 Situation. Increasing possibilities in surgery intervention call for more and higher quality of information. Therefore especially image guided surgery is wanted. The availability of online image-based-information during the operation process can provide a lot of decisions outside of acting-routines. But meanwhile the whole system could be in£uenced. At our department for Human Factors Engineering and Ergonomics participative product development and expert interviews have been established as tools for long-sighted ergonomic product development. Based on an impulse of the medical industry an ORintegrated CT-scan has to be developed by working with participative and strategic development methods. Problem. In surgery we have the situation that information needs to be as complete, as compressed and as real-time as possible, to allow the surgery sta¡ consideration without neglecting the operation routines. While completeness and compression can be solved in most cases by using image-based-information, nowadays the problem of real-time is mainly based on the technical situation in hospitals. This situation is basically described as an architectural and technical separation of OR and patient-scan. The extreme patient movement over high sterility areas, results in process separation of Operation and Scanning. A second problem are the non-interlaced bedding components for the patient which delay the transfer and increase the risk of embedding harms. This situation makes clear, that the development of an online image-guided OR is a task where the whole existing system is in£uenced and might have to be redesigned. Therefore the complexity of the new system expects to take a long period of interdisciplinary development and enclosed the risk to be ``out of date'' when being introduced into clinical routine.

Goal. Development of an integrated system with CT-based image-guided-surgery, anaesthesia and patientscan in one workplace ^ specially focussing on the ergonomic processdemands of the involved sta¡ and patients. This method of development should enable to get all information as complete, as close to the resulting actingprocess and as soon as possible. It should exclude the ``Out of date'' problematic. Therefore a concept is needed, where all involved sta¡ and it's individual task in the period of scanning can be involved as an elementary input for the requirements of the system. To realise a development-routine which excludes the risk of being ``out of date'' a scenario-technique is needed which allows to simulate all interacting tasks of the process together with the involved sta¡ in an empathic manner. Ultimate ambition is a maximum of reality, based on expert knowhow, and a minimum of expense. According to the user requirements future developments in technology and organisation will have to be considered. Solution. An interdisciplinary team of surgery, anaesthesia, medical-engineers, designers and human factors-engineers has been built up, which describes the problem of the future process and to draft guidelines for the expert interviews. Afterwards 21 experts, out of 7 involved disciplines (surgery, anaesthesia, radiology, nursing and hygiene), coming from 9 clinical institutions in Berlin, have been interviewed. Based on this data, we simulate the intraoperative scan together with the involved sta¡ in a scenario orientated way. Based on the results of the interviews a full movable 1 : 5 scale model of the whole system has been realized. This model is accepted by the sta¡ as a scenario-tool for the empathic evaluation. All interactions can be observed and rated by the interdisciplinary team as a control authority for all process related information. In case of variation between expert interviews and user-simulations, the model has to be redesigned and a following empathic round needs to be started. Results. By using this development-method adapted to the technical input of''MAQUET'' (operating tables) the following essential requirements for an intraoperative CT-guided surgery have been evaluated. A process adapted, variable and CT-able patient bedding plain is needed to minimise bedding complications. It is necessary, that the anaesthesia workstation has to be moved in an co-ordinated manner together with the patient which means the same for all medical devices and components. A mechanical separation of bedding plain and arm splints has to be developed according to the di¡erent ergonomic requirements of anaesthesia, surgery and radiology. All moving components should be carried from the ceiling to minimise hygenic problems. And last but not least, while moving the patient, OP-shawls are in contact with unsterile areas, a completely di¡erent way of covering the operation area is needed. The very similar results, comparing expert-interviews and empathic process simulation and the needed time of about 3 months seems to con¢rm the usability of this strategic development tool which can be named as a quick-motion development-process. Dipl. -Ing. Des. Kai^Uwe Neth, Department for Human factors Engineering and Ergonomics ^ Technical University of Berlin

ESCTAIC Abstracts 157

GROUND PLAN OF AN OR AREA ACCORDING TO WORK FLOW REQUIREMENTS - RESULTS OF A STUDY I. Marsolek, W. Friesdorf, Department for Human Factors Engineering and Ergonomics 1999 Situation. An optimized work £ow within the OR area (as the heart of the clinical work system) is a key factor to a productive, e¤cient and quality orientated patient treatment within every hospital. At the Department for Human Factors Engineering and Ergonomics participatory process £ow analysis (focusing equally the involved men, technology and organization) have been established as a basis for the systematical and continuous improvement of work processes within the work system hospital. Problem. The clinical work £ow within every OR area is always in£uenced by its architectural design: therefore the ground plan of an OR area can systematically support the realization of optimized work processes but is also often responsible for making a work £ow improvement impossible. Because of this reason at a very early stage of an architectural OR planning the consideration of optimized clinical work processes (as well for the involved men and technology as for their organization) have to be systematically integrated into the ground plan of the work system OR area. Goal. Our goal is the development of a procedure that enables an interdisciplinary team work for an optimized architectural OR planning based on architectural, economical and engineering knowledge as well as the experience in the every day work processes of the clinical OR sta¡ itself. The external members of the interdisciplinary planning team (e.g. architects, engineers and economists) have to get an intensi¢ed understanding of the work processes within the OR area while at the same time the clinical sta¡ directly involved in the a¡ected work processes needs a systematical support for articulating their problems and ideas of improvement. For this interdisciplinary OR planning process a systematical analysis and planning procedure is needed, that for all members of the interdisciplinary OR planning team brings more transparency to the work £ow within the analyzed OR area, clearly identi¢es all momentarily existing process strengths and de¢cits, and makes all consequences for the work processes within the OR area easily understandable, that will be caused by the realization of any new architectural OR design. Solution. Our solution is the development of a method consisting of two steps: 1. A participatory process £ow analysis of the patient and material £ows within the OR area and other a¡ected clinical departments focussing equally the involved men, technology and hospital organization establishes process transparency: Together with the involved clinical sta¡ the work £ows within the OR area are visualized as process £ow diagrams and transferred into the architectural plans before process strengths and de¢cits are identi¢ed. 2. For transferring those analysis results to the architectural design of the new OR area the identi¢ed patient and material £ows are visualized within the new ground plans as well. Together with all members of the interdisciplinary OR planning team all consequences for the analyzed process £ows (especially all changes for process strengths and de¢cits), that will be caused by the realization of the new architectural OR design, can now be elaborated systematically. Based on this information

an improvement plan for the new architectural design of the analyzed OR area is then elaborated together with all members of the interdisciplinary OR planning team. Results. By using this method within the architectural redesign of the central OR area in the German Red Cross Clinics Ko«penick in Berlin the following architectural improvements were elaborated: A new de¢nition of room utilization (e.g. medical equipment storage rooms are exchanged with material storage rooms) and the arrangement of one main storage room that can be supplied from outside the OR area reduce transportation time and ease the storage inventory. Additional entrances from each OR to adjacent storage rooms reduce transportation time as well and separate the work £ows of the anesthesia and OR team ^ reducing the potential for con£icts. The additional consideration of new (future) technologies (e.g. operation robots) initiated a change in the de¢nition of OR speci¢cations, OR destinations and additional requirements for speci¢c OR entrances to storage rooms, so that an uncomplicated introduction of already planned new medical equipment is now guaranteed. Dipl.-Ing. Ingo Marsolek, Department for Human Factors Engineering and Ergonomics - Technical University of Berlin, Berlin E-mail: [email protected] PREDICTION OF POSTOPERATIVE NAUSEA AND VOMITING USING GENETIC PROGRAMMING P. L. Houweling, A. Koudijs,W. Joosten, Department of Anaesthesiology, Diakonessenhuis, Utrecht,The Netherlands Introduction. In this study we have used a new modelling technique (genetic programming) to predict postoperative nausea and vomiting (PONV). Patients and methods. After institutional approval, the anonymized intra-operative data of 45179 consecutive patients from our automated anaesthesia record database (paradox 4.1, Arkive, San Diego, USA) were converted to comma-separated dataset of 108 variables. This dataset was then randomly divided in a development sample (32389 cases) and a validation sample (12790 cases) and analysed by a genetic programming system (Omega, Cap Gemini, Utrecht, NL) for reliability of behaviour (population analysis) and predictors of nausea and vomiting (data analysis). Variable

Predictive power (%)

Variable

Predictive power (%)

MV time Anaes. Tech. HR ratioa Surgery type Lowest MBP Nitrous oxide Operation room HR variability

70.32 67.30 66.98 65.43 63.35 62.62 61.69 61.01

Lowest HR MBP variability Extubation time Gender Induction time Patient height Surgical time

60.78 60.47 60.24 59.79 59.23 58.15 53.30

Abbreviations: MBP ^ mean blood pressure; HR ^ heart rate; MV ^ mechanical ventilation. a Maximum heart rate divided by the mean heart rate.

158

Journal of Clinical Monitoring and Computing Vol 16 No 2 2000

Results. The genetic programming scheme produced a scoring model for PONV consisting of 15 variables combined with non-linear binary operators to produce a rank-correlation performance above 75. The Table shows the most predictive variables with their individual measured predictive power given in kendall's tau. Conclusion. Although our results are comparable to prospective studies of PONV in large populations [1], this model has to be tested in daily practice: it might be a promising tool as immediate, intraoperative prediction of nausea and vomiting making on-line risk management applicable in anaesthesia. REFERENCE 1. Sinclair DR. Anesthesiology 1998; 89 (3A): A51