Abstracts

SELECTED ABSTRACTS PRESENTED AT THE21ST MEETING OF THE EUROPEAN SOCIETYFOR COMPUTING AND TECHNOLOGY INANAESTHESIA AND INTENSIVE CARE (ESCTAIC)

Amsterdam, The Netherlands, 6th–9th October, 2010

Edited by: A. A. van Dusseldorp, C. Boer, D. S. Karbing,

L. Krummreich, S. E. Rees, S. A. Loer

LIST OF ABSTRACTS

Soraya Abbasi: The Role of Physiological Models inCritiquing Mechanical Ventilation Treatments

Gracee Agrawal: Real-time Detection of Suppression inEEG

Christa Boer: Clinical Experience with PerioperativeNon-invasive Beat-to-beat Arterial Blood PressureMonitoring

Nadja Bressan: Infusion Rate Control Algorithm forTarget Control Infusion using Optimal Control

Chih-Yen Chiang: Rule-based Evaluation for the Patient-controlled Analgesia Clinical Effectiveness

Wolfgang Friesdorf: Professional Design of ClinicalWorking Systems According to Human Factors

Fred de Geus: CAROLA: An Open Source PDMS, After25 Years Still Experimental?

Yori Gidron: The Effects of Stress and HemisphericLateralization on Managerial Decisions

Johan Groeneveld: Value of Central or Mixed Venous O2

Saturation in Guiding Treatment in the Intensive Care UnitGabriel M. Gurman: Professional Stress and the Anes-

thesiologist-how Evident is it?Eliahu Heldman: Salivary Cortisol as a Measure of Pro-

fessional Stress; An Overview and a Description of aStudy with Paramedics

Martin Hurrell: Implementation of a Standards-based,CDA-Compliant Anesthesia Record

Mathieu Jeanne: Analgesia Nociception Index OnlineComputation and Preliminary Clinical Test DuringCholecystectomy Under Remifentanil-PropofolAnaesthesia

Christian Jeleazcov: Pharmacodynamic Modeling ofChanges in Pulse Waveform During Induction of Pro-pofol Anaesthesia in Volunteers: Comparison betweenInvasive and Continuous Non-invasive Measurementsof Pulse Pressure

Pierre Kalfon: Assessing Performances of Glucose ControlAlgorithms on a set of Virtual ICU Patients

Cor Kalkman: Automation and Automation Surprises:Lessons from Aviation. Should Health Care Brace Itself.

Dan S Karbing: Use of the INVENT System for Stan-dardized Quantification of Clinical Preferences TowardsMechanical Ventilator Settings

Talma Kushnir: Moods and Burnout Among Physicians:Associations with Prescribing Medications Communi-cating with Patients, and Referrals for Specialists andDiagnostic Tests

Johannes J van Lieshout: Non-invasive Pulse ContourCardiac Output by Nexfin Technol

Egbert Mik: Monitoring Mitochondrial Oxygenation

Journal of Clinical Monitoring and Computing (2011) 25:3–43

DOI: 10.1007/s10877-011-9276-2 � Springer 2011

Suzani Mohamad Samuri: Absolute EIT Coupled to aBlood Gas Physiological Model for the Assessment ofLung Ventilation in Critical Care Patients

Alan H. Morris: Decision-support for Clinicians-how toImplement

Dick Nickalls: A Linux-based Anaesthesia WorkstationAtsushi Okamura: A Hospital for Post-ICU patients on

Long Term Mechanical Ventilation in JapanAzriel Perel: The Nexfin – a New Non-invasive Monitor

for the Measurement of Continuous Cardiac OutputUlrike Pielmeier: Glucosafe - A Model-based Medical

Decision Support System for Tight Glycemic Controlin Critical Care

Beatrice Podtschaske: Identifying Ergonomic Requirementsof ICT for Healthcare Working Systems

Martyn Read: Safety through standardisationStephen E Rees: The Current Status of the Automatic

Lung Parameter EstimatorNico van Schagen: Components of Anaesthesia Infor-

mation SystemsThomas W L Scheeren: Near Infrared Spectroscopy

(NIRS) to Monitor Tissue Haemoglobin (and myo-globin) Oxygenation

Patrick Schober: Foundation of Tissue Oxygenation:Optimizing Systemic Blood Flow by Trans-oesophagealDoppler (TED) Monitoring

Jan Maarten Schraagen: Human Factors in the ORLothar A Schwarte: Reflectance SpectrophotometryEric Stricker: Simulation in Healthcare - the Techno-

logical PerspectiveSven Zenker: Probabilistic Approaches to solving the

inverse Problem of State and Parameter Estimation forMechanistic Models of Physiology: From Theory toPractice

1. THE ROLE OF PHYSIOLOGICAL MODELS IN CRITIQUINGMECHANICAL VENTILATION TREATMENTS

Tehrani, F. T.1, Abbasi, S.2,3

1Department of Electrical Engineering, California State University,

Fullerton, 800 N. State College Boulevard, Fullerton, California

92831, USA; 2University of Pennsylvania School of Medicine.

Philadelphia, Pennsylvania, USA; 3CHOP Newborn Care at

Pennsylvania Hospital, 800 Spruce Street, Philadelphia, Penn-

sylvania 19107, USA

Introduction: Choosing the correct ventilatory parametersfor ICU patients is an important clinical decision thatneeds to be made with care and sufficient knowledgeabout the patient’s conditions as well as the features oftoday’s advanced ventilators. Computerized decision

support systems can be used as helpful tools in setting theventilator parameters for ICU patients on mechanicalventilation. There have been many decision support sys-tems developed by various researchers for this purposeover the past few decades [1, 2]. While decision supportsystems can be used by clinicians to choose the ventilatoryparameters, a system based on a physiological model of thepatient can provide further advice to clinicians by pre-dicting the treatment outcome and critiquing their deci-sions. The present study was designed to examine theeffectiveness of a model based critiquing system.Methods: The system used to critique the ventilatorytreatment options in this study is based on an earlierphysiological model of the infant respiratory system [3].That model consists of a continuous plant and a discretecontroller. For the purpose of this study, the discretecontroller of the model was replaced by a positive pressuremechanical ventilator providing pressure to the infant’sairways and the inspiratory gas. A block diagram of thissystem is shown in Figure 1. As shown in this figure, thesystem includes lungs, body tissue and brain tissue. Thelung volume is continuously time varying and the effect ofshunt in the lung, changes in cardiac output, and thearterial transport delays are included in the model. Themass balance equations of these compartments are pro-vided in Reference 3. The inspiratory gas is provided by apositive pressure mechanical ventilator to the lungs. Theexpansion of the lungs is controlled by the amount ofpressure applied by the ventilator and the infant’s lungmechanics. The inspiratory gas comes into contact withthe alveolar tissue. The venous blood supplied to the lungsby the heart absorbs oxygen from the inspired gas andloses its carbon dioxide to it during inspiration. The gas isthen exhaled and the oxygenated blood leaves the lungsbut mixes with some venous blood due shunt in the lungbefore being pumped by the heart and delivered to the

Fig. 1 A block diagram of the model.

4 Journal of Clinical Monitoring and Computing

brain and the body tissues. A list of the internal parametersof this model and their default values are provided inReference #3. For the purpose of this study, the venti-latory and physiological data of an infant reported in aprevious study [4] were used. In that study, a computer-ized system for mechanical ventilation called FLEX wasused to determine the optimal ventilatory parameters of agroup of infants. FLEX is a new system that can be used asa closed-loop controller as well as an open-loop decisionsupport advisory system for mechanical ventilation. FLEXincludes the features of a patented commercial ventilatorymode called Adaptive Support Ventilation (ASV) [5].FLEX further includes many additional features forcontrol of fraction of inspired oxygen (FIO2), positive

end-expiratory pressure (PEEP), minute ventilation, andweaning. More details of FLEX can be found in otherreferences [4] and are not repeated here for brevity. Theinfant whose data is used in this study is infant #5 inReference #4, who is a male one-day old infant of 2.5 kgweight, with respiratory distress syndrome (RDS). Theclinician’s set of ventilator parameters as well as the rec-ommended parameters by FLEX were input to thephysiological model of Figure 1 in two separate simula-tion studies. The simulation results were used in a com-parison between the two treatment options.

Results: The clinician’s set of ventilatory parameters forthe infant of this study resulted in a tidal volume of5.7 ± 2.7 ml, and a breathing rate of 68 ± 18 breaths/minincluding an intermittent mandatory (IMV) rate of 35breaths/min. FIO2 was set at 21%, and PEEP was 5cmH2O. The respiratory airway resistance and respiratorydynamic compliance of this infant were measured at143 ± 60 cmH2O/l/s, and 0.93 ± 0.39 ml/cmH2Orespectively. The FLEX computerized system recom-mended a ventilation of 0.66 l/min, a total respiratory rateof 45.5 breaths/min including the IMV rate, an FIO2 of21%, and a PEEP of 4.2 cmH2O. Figures 2a and 2b showthe simulation results of arterial partial pressures of CO2

(PaCO2), and O2 (PaO2

) for this infant by using the clini-cian’s set of parameters and the FLEX recommendedparameters, respectively.

Discussion: According to the simulation results ofFigure 2a, the use of the clinician’s set of parametersshould result in hypercapnia with PaCO2 rising to about47 mmHg and a decline in PaO2

. At the next round ofevaluation in Reference #4, the end-tidal CO2 pressureof this infant was measured at 43 mmHg, representingmild hypercapnia, and the arterial oxygen saturationof this infant was somewhat decreased. The results ofFigure 2b predict that by using the FLEX recommendedvalues, there would not be any hypercapnia or anyreduction in PaO2

. These results show that by using themodel based critiquing system, the clinician would havebeen able to get more information about the treatmentoutcomes and make a more informed choice betweennormocapnia and mild permissive hypercapnia for thisinfant. This example indicates that systems based onphysiological models have the potential to be used ashelpful tools in critiquing ventilatory treatment options.

REFERENCES

1. Sittig DF, Gardner RM, Morris AH, Wallace CJ. Clinicalevaluation of computer-based respiratory care algorithms.Int J Clin Monit Comput 1990; 7(3): 177–85.

Fig. 2a Simulation results by using the clinician’s set of ventilatoryparameters for infant #5 in Reference #4.

Fig. 2b Simulation results by using the ventilation parameters recom-mended by a computerized system called FLEX, for infant #5 in Reference#4.

Abstracts 5

2. Tehrani FT, Roum JH. Intelligent decision supportsystems for mechanical ventilation. Artif Intell Med2008; 44(3): 171–182.

3. Tehrani FT. Mathematical Analysis and ComputerSimulation of the Respiratory System in the NewbornInfant. IEEE Transactions on Biomedical Engineering1993; 40(5): 475–481.

4. Tehrani FT, Abbasi S. Evaluation of a ComputerizedSystem for Mechanical Ventilation of Infants. Journal ofClinical Monitoring and Computing 2009; 23: 93–104.

5. Tehrani FT. Automatic Control of Mechanical Ven-tilation. Part 2: The Existing Techniques and FutureTrends. Journal of Clinical Monitoring and Comput-ing 2008; 22: 417–424.

2. REAL-TIME DETECTION OF SUPPRESSION IN EEG

Gracee Agrawal

NeuroWave Systems, Inc., Cleveland Heights, Ohio, USA

Introduction: In clinical settings, timely and robust detec-tion of suppression in EEG signals is very important. Inparticular, the presence of burst-suppression pattern inEEG indicates deep general anesthesia or underlying se-vere neuropathology such as stroke and ischemia. Withthe advent of neurological monitoring for anestheticmanagement, anesthesiologists consider the presence ofsuppression as an ominous sign when not related to a highdosage of anesthetic drugs. The automatic detection ofsuppression in EEG waveform is typically carried outbased on a method inherited from visual observation,where suppression is defined as a period longer than 0.5 swith peak-to-peak amplitude less than 5 lV pp [1].However, being based on peak-to-peak measurements,this method is particularly sensitive to noise and can fail todetect suppression in certain conditions. In this study, weinvestigate the changes in the first derivative of EEGsignals during suppression periods to develop an algorithmfor automatic detection of suppression in real time. Wealso provide preliminary results where we compare theperformance of our novel method to the visual annota-tions by an expert technologist. We further provide acomparison with the traditional peak-to-peak method forreference.Methods:(a) EEG recordingsRetrospective analysis was done on EEG signals acquiredfrom 5 patients undergoing cardiovascular surgeries. TheEEG signal was obtained using bilateral EasyPrepTM

electrodes connected to the NeuroSENSETM monitor at asampling frequency of 900 S/s.

The suppression periods in the EEG signal were man-ually annotated. The EEG data, with the two channelscombined together, consisted of a total of 5.19 h ofannotated suppression periods.

(b) First derivative of EEGThe first time-derivative of EEG signal has shown somepromise as an easily measurable proxy order parameter forthe cerebral cortex [2]. The time-varying EEG signalsreflect the fluctuations in the soma potential, and the firstderivative of these fluctuations has been shown to bestrongly linked to the mean soma potential.

A burst-suppression pattern in EEG may be generateddue to activation of large and small number of neurons inan alternating fashion. Hence, it is expected to beaccompanied by waxing and waning of the mean somapotential.

Figure 1 presents a sample EEG segment exhibitingburst suppression pattern (1a) along with its first time-derivative (1b). One interesting thing to note is that thedetection of suppression periods from the first derivativeof EEG signal looks similar to the problem of silencedetection in speech signals, or the detection of muscleinactivity in EMG signals.

For the time-varying EEG signal s(t), we postulate thatthe median absolute value of the first derivative, i.e.,median(|ds/dt|), should be a good candidate for sup-pression detection. This is because it gives a robust mea-sure of the rate of change of EEG while reducing theimportance of outliers such as ECG artifacts, etc.

(c) Suppression DetectionAll the data processing was done using Matlab� v.7.4.0.The EEG was first preprocessed using the NeuroSENSEalgorithm v.2.1.1.2 (notch filter at 50/60 Hz, high-passfilter at 0.5 Hz, decimation to 128 S/s) [3]. Suppressionperiods were then detected from the EEG data using boththe traditional method based on peak-to-peak amplitude,and the new proposed method based on the first time-derivative of the EEG signal. Figure 1 shows the classifi-cation of a sample EEG segment into burst and suppressionby visual detection (1c) and automatic detection usingboth the algorithms (1d, 1e).

Results:The ROC curves for the detection of suppression periodsin EEG signal using both the traditional and new methodare shown in Figure 2. It is apparent that the new methodperforms better than the traditional method in thedetection of suppression. Using the traditional methodwith a detection threshold of 5 lV pp, the specificity andsensitivity are 90.60% and 80.52% respectively. For thesame specificity, our proposed method gives an increasedsensitivity of 85.73%.

6 Journal of Clinical Monitoring and Computing

Conclusions & Discussion:This paper proposes a new method for the automaticdetection of suppression in EEG signals based on its first-time derivative. The preliminary results obtained with thismethod were compared with the visual annotations by anexpert technologist and the traditional peak-to-peakmethod. The new method appears to perform better thanthe traditional method in these clinical cases.

A more challenging dataset containing near-suppressionpatterns, for which the traditional peak-to-peak methodmay perform poorly, is currently being investigated.

REFERENCES

[1] Rampil IJ, Lockhart SH, Eger EI, Yasuda N, WeiskopfRB & Cahalan MK. The electroencephalographiceffects of desflurane in humans. Anesthesiology. 1991;74:434–439.

[2] Sleigh JW, Steyn-Ross DA & Steyn-Ross ML. Thefirst time-derivative of the EEG: A possible proxy forthe order-parameter for the cerebral cortex. ProcComplex Systems. 1998;74–79.

[3] Bibian S & Zikov T. NeuroSENSE� Monitorwith WAVCNS Cortical Quantifier: A DeterministicApproach to EEG Analysis. White Paper. www.neurowavesystems.com.

3. CLINICAL EXPERIENCE WITH PERIOPERATIVE NON-INVASIVE BEAT-TO-BEAT ARTERIAL BLOOD PRESSUREMONITORING

Dr. Christa Boer

Department of Anesthesiology, Institute for Cardiovascular Research

Vrije Universiteit, VU University Medical Center, Amsterdam,

The Netherlands

Increasing age and preexisting comorbidities like hyper-tension and diabetes characterize the current population ofpatients undergoing anesthesia and surgery. Among oth-ers, this may increase the risk for hemodynamic instabilityand the development of complications in the periopera-tive period. Consequently, a shift in focus has beenobserved from intraoperative care towards preoperative riskassessment and postoperative monitoring in these patients.

The increasing attention for preoperative risk assess-ment, diagnosis of hemodynamic abnormalities and earlydetection of postoperative complications in the generalsurgical population warrants non-invasive hemodynamicmonitoring techniques that allow evaluation of the car-diopulmonary condition of the patient. In particular,indices like pulse pressure variation and pulse contour-derived cardiac output require beat-to-beat blood pressuremeasurements, which are not provided by routine bloodpressure manometers. Moreover, most beat-to-beat bloodpressure measurement methods are limited by their invasivenature, and thus restricted to patients with an intra-arterialentrance like patients admitted to an intensive care unit.

Fig. 1 (a) Sample EEG segment; (b) First derivative; (c) Visual detection;(d) Automatic detection using traditional method; (e) Automatic detectionusing new method [s: suppression; b: burst].

Fig. 2 ROC curve for suppression detection for the traditional and newmethod.

Abstracts 7

We here present our clinical experience with a non-invasive, beat-to-beat arterial blood pressure monitoringdevice (Nexfin HD) during preoperative risk assessment andintraoperative and postoperative hemodynamic monitoring.Using a finger blood pressure cuff, this device enables arterialblood pressure measurements in adults and potentiallychildren. Moreover, Nexfin HD allows derivation of dy-namic preload variables like the pulse pressure variation andcardiac output, which may be useful to evaluate the peri-operative circulatory state of surgical patients.

Data will be shown with regard to the reproducibilityof non-invasive arterial blood pressure measurements andthe comparability with invasive arterial blood pressuremeasurements in adults and children. Additionally, cardiacoutput monitoring by Nexfin HD and echo-Doppler iscompared, and its application in pulse pressure variationmonitoring will be presented. Finally, data will be shownwith regard to Nexfin HD non-invasive arterial bloodpressure measurements during preoperative evaluation ofautonomic function, postoperative monitoring of car-diopulmonary interactions and intraoperative monitoringof blood pressure and cardiac output.

Our clinical experience with Nexfin HD provides arationale for non-invasive, beat-to-beat arterial bloodpressure measurements as valuable alternative for hemo-dynamic monitoring in the perioperative period in thegeneral surgical population.

4. INFUSION RATE CONTROL ALGORITHM FOR TARGETCONTROL INFUSION USING OPTIMAL CONTROL

N. Bressan, Miguel Pinto, Heber Sobreira,

P. Amorim§, C. S. Nunes*, A. Paulo Moreira

Faculdade de Engenharia da Universidade do Porto, Porto, Portugal

Introduction: Target Controlled Infusion (TCI) is a tech-nique used in Total Intravenous Anaesthesia (TIVA)practice. TCI system allows the anaesthesiologist target adrug concentration in a specific site remotely command-ing an infusion device. TCI system incorporates: Phar-macokinetic/Pharmacodynamic (PK/PD) Model toestimate the concentration at plasma and site-effect of thedrug and; an Infusion Rate Control Algorithms (IRCA)to control the drug dose to be titrated by infusion pumps.The IRCA is based on BET (Bolus, Elimination andTransference) scheme [1] describe by a bolus and a vectorof infusions. The bolus is a fast infusion to fill the centralcompartment followed by the vector of infusions tocompensate the elimination between compartments andthe transfer between peripheral compartments, aimingreach and maintain the target concentration. This work

presents a new controller (NC) developed to controlplasma concentration (Cp) and effect-site concentration(Ce), to be used in our TCI systems Anaesthesia Syn-chronization Software (ASYS) [2], as well an improve-ment of this NC using optimal control.

Method: The algorithm was developed in Matlab beenimplemented and test in simulation mode with ASYSimplemented in LabView. This study is based on the ana-lytical solution of Shafer and on the algorithm of Poucke.[3; 4] From the effect-site compartment only the ke0 rateconstant was considered. The NC was develop, based onthe continuous model, with a matrix based control rule:

_XðtÞ ¼ A � XðtÞ þ B � uðtÞ; YðtÞ ¼ C � XðtÞ

Where:

A ¼

�ðK10þK12þK13Þ K21 K31 0

K12 �K21 0 0

K13 0 �K31 0

Ke0 0 0 �Ke0

26664

37775

B ¼

1

0

0

0

26664

37775 C ¼ 1 0 0 0½ �

Below the conversion of the continuous model in spacestates to discrete where Cp is the output and the input thedose (i). Equating the state space model in order to theinput it is possible to know the exact value of the drugamount to obtain the next plasma concentration (Cpt):

DoseðtÞ ¼ 0:36 � ðC � AÞ�1ðCpt � 1000 � V1 � C � A � XÞ;V1 ¼ PatientWeight � Vc ð1Þ

8 Journal of Clinical Monitoring and Computing

where kn and Vc are pharmacokinetic parameters. TheNC, represented at diagram block Figure 1, is mainlycomposed by equation 1 following the BET schemeautomatically. The gain K, Figure 1, was implemented toavoid the overshoot with optimal steeling time using theoptimal control technique. The block diagram is repre-sented by equation 2, as follow:

Cpt ¼ K � ððCet � Xð4ÞÞ=VcÞ=1000þ Cet ð2Þ

To solve the problem using optimal control the mathe-matical model was wrote in AMPL (A MathematicalProgramming Language) [5] allowing write a set of con-straints and the cost function that is pretended minimize.Once KNITRO [6] solver supports AMPL was used tosolve the optimal control problem.

To test the new controller (NC) a simulation protocolwas implemented with Marsh’s Pk model for propofol and70 kg male patient. The experimental protocol used was:an initial propofol target for Cp and Ce set at 3 lg/ml andkept until steady state was reached at which point it waschanged to 5 lg/ml and kept until steady state was reached.

Results: The NC demonstrated a good response instationary state, figure 1. However transitory phase for thetarget of effect-site concentration (Cet) at 3 lg/ml shownan overshoot, not viewable for Cet = 5 lg/ml or Cpt.Aiming to eliminate the overshoot consecutively thedrug toxicity and the overdose (plasma concentration),identified the optimal control technique was implementedin the NC. The NA using optimal control introduceda constraint Cplim, Figure 1. The response obtained,Figure 1, demonstrated the expected result.

Discussion: The results with the NC provide enougharguments of a robust and reliable IRCA. This studypresent an algorithm mathematically simpler than most ofthe existing algorithms and the results shown that it per-formance accurately. We believe that the fact it is simplermay offer advantages when used in the clinical setup tocontrol infusion devices; more tests will be needed to fullyassess it and a future work will present the optimal controlusing as tune the Lean Body Mass (LBM) and the target ofeffect-site.

REFERENCES

1. Schwilden H., European Journal of Clinical Pharma-cology, vol. 20, 1981, pp 379: 386

2. N. Bressan et al., IEEE Transactions on Biomedical,2008, pp 5543: 47

3. Shafer S.L., Journal of Pharmacokinetics and Bio-pharmaceutics, vol. 20, 1992, pp 147: 169

4. Poucke Van E.G., IEEE Transactions on BiomedicalEngineering, vol. 51, no 11, November 2004

5. Access at March/2010: http: //www.ampl.com [6]Access at March/2010: http: //www.ziena.com/knitro.html.

5. RULE-BASED EVALUATION FOR THE PATIENT-CONTROLLED ANALGESIA CLINICAL EFFECTIVENESS

Chih-Yen Chiang, Hung-Chun Chen,

Kuang-Yi Chang*, Mei-Yung Tsou*,

I-Ting Kuo, Steen J. Hsu�, Chia-Tai Chan

Institute of Biomedical Engineering, National Yang-Ming

University, Taipei, Taiwan; *Department of Anesthesiology,

Taipei Veterans General Hospital, Taipei, Taiwan; �Department

of Information Management Minghsin University of Science and

Technology, Hsinchu, Taiwan

Introduction: The self-regulated patient-controlled analge-sia (PCA) has become an established procedure for clinicalpain relief through the electric-mechanical pumpingcontrol to deliver a bolus dose and a background infusion.Since the pain is subjective sense and pain measurementrelies on the verbal report of patients, it is difficult toquantify and evaluate the clinical effectiveness of PCAtherapy.1 The PCA logs that stored the drug deliverysettings, patients’ demand/delivery condition, and thera-peutic history were often neglected by most hospitals dueto the shortage of manpower. Therefore, a rule-basedevaluation of clinical effectiveness for miscellaneous PCAprotocols using a novel index based on Fuzzy logic modelis proposed to evaluate PCA pain care quality. Firstly, theproposed system intends to explore the relationshipbetween patients’ demand and analgesic delivery (termedD/D ratio) during the therapy. Through a fuzzy-ruledclassification and discrimination, a referable PCA clinicaleffectiveness index that reveals therapeutic effectiveness ofPCA therapy will be provided to the anesthesiologists torealize the patient’s pain relief profile and recovery status.

Methods: Inpatients who adopted the PCA therapyassumed the pain shall be adequately relieved. However,

Fig. 1 Propofol Cet with optimal control and block diagram with newcontroller.

Abstracts 9

many still experienced certain level of pain during thetherapy that resulted in the decline of patient’s postop-erative recovery and dissatisfaction about PCA service. Inorder to realize the pain relief profile during the PCAtherapy, the analgesic demand and delivery events withcorresponding timestamps were extracted from the PCAmicrocomputer log. The D/D ratio which equals to thedemand counts divided by the delivery counts was used asa major input. The time series of D/D ratio will be pre-processed through the aggregation technique of slidingwindow. A 4-h window is adopted to calculate the D/Dratio and the window slides every 1 h. This means a 72-h

PCA therapy will be aggregated into 69 dataset of D/Dratio. During our previous study in 2006� 2009, theproposed PCA EDC (Electronic Data Collection) systemhad collected over 16,000 patients’ clinical data.2 Patientswho adopted orthopedic surgery and IV-PCA therapywith constant 0.1–0.2 ml/hr background infusion wereselected as samples in this study. In order to obtain areasonable PCA clinical effectiveness index, we chose fivelinguistic levels of the D/D ratio as fuzzifier inputs. Theyare Increase Big (IB), Increase Small (IS), Zero (Z),Decrease Small (DS), and Decrease Big (DB) respectively.In addition, each level is further classified into three cat-egories with varying extents: Big, Medium, and Small.The fuzzy inference engine and the defuzzifier werecooperated with opinions of the anesthesiologists. Thegenerated PCA clinical effectiveness index will give asuitable appraisal after the PCA therapy, and the overviewis illustrated in Figure 1.

Results: The statistics of D/D ratio patterns collectedfrom a 4-h window that slides every 1-h is shown inTable 1. The D/D ratio profile of a 48-h PCA therapy canbe classified into 6 patterns: (1) Flat profile; (2) Descendingprofile; (3) Middle-High profile with peak in 12� 24 htherapy; (4) Middle-High profile with peak in 24–36 htherapy; (5) Ascending profile; (6) M-Peak profile.

The PCA clinical effectiveness index graded the PCAtherapeutic satisfaction into 5 levels, and they are ‘‘Best’’,‘‘Good’’, ‘‘Normal’’, ‘‘Bad’’, and ‘‘Worst’’, respectively.The statistics is drawn in Figure 2. The PCA satisfaction is79% that summed the level of ‘‘Best’’, ‘‘Good’’, and‘‘Normal’’. The result can faithfully reflect the PCA sat-isfaction from the questionnaire of the PCA nurses’ dailyvisitation.

Fig. 1 PCA clinical effectiveness index is generated to give appraisal afterPCA therapy. Fig. 2 The satisfaction statistics of PCA clinical effectiveness index.

Table 1 Six Patterns of D/D ratio based on sliding window

No. Pattern type Small

amount

Percent

1 Flat profile: normally

distributed (D/D ratio = 1�2)

30 31

2 Decending profile:

peak lies in 1–2 h

28 29

3 Middle-high profile:

peak lies in 12–24 h

19 20

4 Middle-high profile:

peak lies in 24–36 h

7 7

5 Ascending-high profile:

peak lies in 36–48 h

8 8

6 M-peak profile: multiple

peaks

5 5

Summation 97 100

10 Journal of Clinical Monitoring and Computing

Discussion: Essentially, the best satisfaction of PCAtherapy lies in the D/D ratio approaching 1 because onedemand earns one delivery of drug injection and theanalgesic effect happens to relieve patient’s pain.3 Mostorthopedic postoperative patients were arranged to takeoff-bed rehabilitation in the second day after surgerywhich led to the middle-high profile of D/D ratio.

The PCA clinical effectiveness index and the pain reliefprofile offer doctors important treatment reference, espe-cially for the patients with second or recurrent surgery. Inaddition, the pain relief profile on different divisions ofmedicine and various surgical sites are classified and sum-marized into useful information for doctors. However, theassessment of PCA clinical effectiveness exploited the off-line data mining methods to evaluating the clinical effec-tiveness. An on-line PCA effectiveness evaluation is goingto be planned to perform better therapeutic quality.

REFERENCES

1. I-Ting Kuo, Chih-Yen Chiang, Kuang-Yi Chang,Mei-Ling Yeh, Steen J.Hsu, Chia-Tai Chan, Designand Implementation of PCA Dosage Information Generator.Journal of Clinical Monitoring and Computing, 2009.24: p. 20–22.

2. I-Ting Kuo, et al., Web-Based Electronic Data CollectionSystem to Support Patient-Controlled Analgesia in Taiwan.The Clinical Journal of Pain, Journal of ClinicalMonitoring and Computing, 2009. 24: p. 22–24.

3. Chang, K.-Y., et al., Determinants of Patient-controlledEpidural Analgesia Requirements: A Prospective Analysis of1753 Patients. The Clinical Journal of Pain, 2006. 22(9):pp. 751–756. 10.1097/01.ajp.0000210924.56654.03.

6. PROFESSIONAL DESIGN OF CLINICAL WORKING SYSTEMSACCORDING TO HUMAN FACTORS

Friesdorf, W., Podtschaske B., Gartner J.,

Aciksoz-Tavasli, F.

Department for Human Factors Engineering & Ergonomics;

Technical University of Berlin

Introduction: Clinical working systems (CWS) are underpressure: patients are increasingly informed and demand-ing, critical treatment incidents are under public discus-sion, resources are limited, and, as a result, clinicians areover-challenged. Their valuable intrinsic motivation isgetting lost by stressing working conditions. CWS have

been organically evolved; no doubt a lot of optimizationpotential can be presumed. A sustainable re-engineeringaccording to human factors is required. But how toproceed and who is in charge?

F. W. Taylor has defined ‘‘Principles of ScientificManagement’’ for industry almost a century ago [1]. Theapplication of these principles to CWS seems to beobvious but must fail due to two strong reasons: 1) patienttreatment is a complex task with severe limitations forstandardization and work flow planning; 2) the work isperformed by experts, who deal with complexity, notcomparable with workers at an assembly-line. But if a‘‘simple’’ industrial working system requires scientificmanagement; this must be even more necessary in com-plex CWS. Medicine is using scientific methods todemonstrate the effectiveness of new procedure or drugs.Curiously there are no established methods to measure,prove or compare the quality and the efficiency of clinicalworking processes (HOW is the medical procedure per-formed?) [2].

Starting with Taylor’s principles a lot of research wasdone leading to a human centered scientific management.A lot of knowledge and experience has been achieved inseveral domains e.g. in manufacturing, aerospace, agri-culture, transportation, power supply. Health care is still atthe beginning. Today we see several different scientificcommunities dealing worldwide with humans & work:Ergonomics, Human Factors, Human Factors Engineer-ing, Human Computer Interaction, and others withoverlapping scopes. We prefer Ergonomics, which pro-vides the comprehensive consideration of human, tech-nology and organization:

‘‘Ergonomics (or human factors) is the scientific disci-pline concerned with the understanding of the interac-tions among humans and other elements of a system, andthe profession that applies theoretical principles, dataand methods to design in order to optimize human wellbeing and overall system performance. Practitioners ofergonomics, ergonomists, contribute to the planning,design and evaluation of tasks, jobs, products, organiza-tions, environments and systems in order to make themcompatible with the needs, abilities and limitations ofpeople.’’ [3]

Goal: Ergonomics/human factors as a scientific disci-pline should understand the special requirements of apatient treatment and should supplement theories, meth-ods, and tools accordingly. Hospitals, industry (medicalproducts) and especially medical societies should make useof ergonomics competence. Yoel Donchin has coined theterm ‘‘Medico-Ergonomics’’, which describes the crosssection between medical needs, technical possibilities, andergonomic rules excellently [4]. The goal is to vitalizeMedico-Ergonomics.

Abstracts 11

Concept: Within simple working systems an externalcounselor (e.g. an ergonomist) can analyze working pro-cesses and give advice for optimization. It is easy for her/him to understand the given production-line as well as thetechnical and organizational options. In comparison thecomplexity of a clinical working system is rather difficultto understand by counselors without medical background(expert system). This does not mean, that a counselormust be trained in different disciplines (interdisciplinarybackground), but rather that Medico-Ergonomics requiresa co-operation between disciplines (multidisciplinaryconcept) as well as between structures; all stakeholdersshould be involved. A systems approach with three levelscould support the selection of stakeholders.

(1) Micro-Ergonomics comprises the interactions betweenthe system elements: patient, clinicians (physician,nurses, etc.), and technology (medical products).

(2) Process-Ergonomics is concerned with treatment tasksperformed on a time axis; activities and interactionsfrom the micro-level are done in serial and in parallelto complete a given task. This level includes theorganization of work flow.

(3) Macro-Ergonomics covers the hospital as an institutionwithin a health care market (external view) and withall aspects of running a human centered company(internal view). Table 1 shows groups of stakeholdersand their influence/contribution.

The counselor gets into the coordination role, takingcare for the communication between the stakeholders andthe integration of their knowledge [5].

Realization: For implementation a business model hasbeen developed named ‘‘Fabrica medica’’, including aproject format with the following steps: 1) Focus defini-tion; 2) Stakeholder selection; 3) Three workshops a)

Synchronization (shared mental model), b) Fabrication(multidisciplinary solution), c) Validation; 4) Exploitation.The time frame is 9 to 12 months per project. The firstprojects have been defined and are in progress.

REFERENCES

1. Taylor, F. W. (1911). The Principles of ScientificManagement. New York: Harper Bros.

2. Marsolek, I. & Friesdorf, W. (2006). Work Systemsand Process Analysis in Health Care. In: Carayon, P.(Ed.): Handbook of Human Factors and Ergonomicsin Healthcare and Patient Safety. Mahwah, New Jer-sey: Lawrence Erlbaum Associates, 649–662

3. Council of the International Ergonomics Association(IEA) (2000). www.iea.cc/browse.php?contID=what_is_ergonomics; access April 10th. 2010

4. Donchin, Y. (2005). Medication Errors: Good tips forprevention. European Society of Intensive CareMedicine (ESICM), Congress Amsterdam

5. St. Steinheider, B., Burger, E. (2000). Kooperation ininterdisziplinaren Entwicklungsteams. In GfA e.V.(Hrsg.), Komplexe Arbeitssysteme – Herausforderungfur Analyse und Gestaltung. (46). Dortmund, GfA-Press, 553–557

7. CAROLA: AN OPEN SOURCE PDMS, AFTER 25 YEARS STILLEXPERIMENTAL?

de Geus, Fred, MD

Information Scientist of the Department of Anesthesiology,

State University Hospital of Groningen

In 1983 the first Carola PDMS system was introduced forThoracic surgery ORs. After 25 years it is still operationalin all Thoracic OR’s at the University Medical CenterGroningen the Netherlands and the Kerckhoff Klinik inBad Nauheim, Germany. Since its inception it has beenunder constant development, adjusting to the needs andwishes of its users and the changing landscape in computersoftware and hardware. Carola PDMS data from morethan 100,000 thorax operations (since 1985) are accessiblethrough a web interface.

• What is the secret to Carola’s longevity?• What can it do what other PDMS systems can’t?• Is comparable software available from commercial

vendors?• What are the advantages of open source software?

Table 1 Optimization potential and influence/contribution of stakeholders based on our experience

System

level

Optimization

potential

Clinicians1) Hospital

OPERATORS

Industry Politics

Macro-

Ergonomics

++ ++ +++ +2) +++

Process-

Ergonomics

+++ +++ ++ +++ +3)

Micro-

Ergonomics

+ +++ + +++ 0

1)representative for physicians, nurses and other clinical experts, including the patients’

requirements.2)hospital information systems should support the management.3)regula-

tions/incentives for integrated treatment solutions.

12 Journal of Clinical Monitoring and Computing

In the introduction the goals and design principlesbehind Carola will be highlighted. Using real-world-examples, it will be described how Carola is used in theOR, how it communicates with the user, how equipmentis interfaced, and how data are imported from and exportedto external systems. A short description of how a Carolasystem can be set up will also be given.

A Carola satellite system has also been developed fortechnicians operating a heart lung machine. Data sharingin the OR between it and the Carola Anesthesia systemwill be described.

PDM Systems are an essential part of medical care in theinformation age, but they can be complex and extremelycostly. Not only purchase costs need to be considered, butalso the costs of maintenance, software and hardware up-dates and new developments need to be considered. TheCarola approach, using a stable open source platform,attention to software and hardware portability have keptthese costs low, while at the same time allowing theflexibility to adapt to 25 years of changing hardware,software and evolving user expectations. Examples of howCarola can be adapted to some expected new develop-ments in the coming years will be described.

8. THE EFFECTS OF STRESS AND HEMISPHERICLATERALIZATION ON MANAGERIAL DECISIONS

Yori Gidron, PhD

There is a large accumulation of scientific research on themodels, processes, underlying biology, health conse-quences and management of job stress. Studies have shownthat stress influences decision making in interesting ways,by not simply affecting it quantitatively but rather quali-tatively. Under stress, people make different types ofdecisions and consider fewer options. Furthermore, duringstress, people shift to use more the right hemisphere, whichis more global and ‘‘intuitive’’, relatively abandoning themore analytic and ‘‘logical’’ left hemisphere. People canalso be classified as having relatively more left versus righthemispheric lateralization (HL). Would may be the con-sequences of the HL especially during stress? I will present aseries of findings we recently discovered in business stu-dents and managers, showing that both stress and HLinteract in surprising ways. These findings theoreticallymay be manifested in physicians as well, and could haveimportant implications for decision making at criticaltimes. In some studies, we show that it is the left HL peoplewho may ‘‘pay a price’’ and become less analytic, while inother times, it is the right HL people who seem lessimmune to effects of stress. This of course also raises thequestion whether analytic or intuitive thinking are correct

at all times? These issues will be presented and discussed inmy talk, and will be applied to medical scenarios.

9. VALUE OF CENTRAL OR MIXED VENOUS O2 SATURATIONIN GUIDING TREATMENT IN THE INTENSIVE CARE UNIT.

ABJ Groeneveld

Dept of Intensive Care Medicine, VU University Medical Center,

Amsterdam, The Netherlands

Since the publication by Rivers et al. on the value of earlygoal-directed therapy for septic shock, utilizing amongothers the central venous O2 saturation to guide treatmentand showing a survival benefit of such policy, the use ofO2 saturation in venous blood has been propagated toguide the hemodynamic management of a variety ofcritical disorders. This can be done with and withoutsupplemental cardiac output measurements. Pitfallsinclude the disparity between central and mixed venousO2 saturation depending on varying admixture of venousblood including that from the coronary sinus, particularyin shock. Although a central or mixed venous SO2 below70 and 65%, respectively, may denote inadequacy of O2

delivery for tissue needs, the adequacy of cardiac output isonly indirectly assessed, and should be supplemented byother signs of organ malperfusion including lactic acide-mia and oliguria. The value of (continuous) measurementsin predicting fluid responsiveness of the heart, i.e anincrease in cardiac output with fluid loading, is currentlyunder investigation. A low saturation could be helpfulwhen cardiac function is relatively normal but whencardiac failure or cardiogenic shock supervenes. Thetherapeutic consequences, in terms of fluid loading orinotropic therapy, of relatively low saturations thus remainunclear. In the mean time the saturations are used at thebedside in the intensive care unit to supplement otherobservations in judging the adequacy of cardiac output.

10. PROFESSIONAL STRESS AND THE ANESTHESIOLOGIST:HOW EVIDENT IS IT?

Gabriel M. Gurman, M.D.

Professor of Anesthesiology and Critical Care, Ben Gurion

University of the Negev, Beer Sheva, Israel; gurman@bgu.ac.il

Anesthesiology, a stressful professionThere is no doubt that Anesthesiology as a profession is asource of stress for the physician.

In spite of the tremendous reduction in the rate ofanesthesia morbidity and mortality, due to new tech-nology, new drugs and higher educated professionals,

Abstracts 13

Anesthesiology is becoming in the new millennium moreand more dangerous for its practitioners.

Anesthesiology is a stressful profession because amultifactorial etiology. One can only remember the factthat this is a service profession, dealing with a ‘‘tem-porary pharmacological intoxication, using many blindmethods and producing from time to tome complica-tions difficult to accept by the patient of family, tounderstand that our daily activity is full of stressfulevents.

The stress of the anesthesiologist is a chronic one, butmany adverse events can superimpose and accentuateacutely the psychological situation of the professional inthe operating room or in the critical care environment(acute stress): a sudden deterioration of the patient sta-bility, altercations with peers, etc.

The impact of stress on the anesthesiologist well beingis obvious. It includes drug abuse, a high rate of suicide,the need for early retirement, and the burnout syndrome.

Some earlier studies indicated that anesthesiologistsdevelop malignancy in a higher rate than other profes-sional groups, but lately it was shown that our averagelength of life increases all the time (Kats and Slade, J ClinAnesth 2006;18: 405).

But even so, our quality of life is too often poor. Wesleep less, we are overworked, we are very tired and whenthe fatigue conquers our brain, our ability to judge andtake decisions is affected.

In a recent study done on 39 anesthesiologists (Gur-man, Gidron, Gurski, unpublished data) we could provethat most of the subjects were found be stressed (level ofsalivary cortisol) and with a tendency to burnout.

Remedies to be taken into considerationDuring the last years some authors concentrated on

possible advises to be followed in order to improve notonly our length of life but also is quality.

Since the etiologic factor of well being impairment isthe stress (with its direct effects on cardiovascular system,nervous system AND coagulation) some remedies havebeen proposed for reducing the level of professional stressamong anesthesiologists.

This is just a short list of proposals:

*early detection of stressed colleagues*improving of human engineering to appease our daily

activity*education and self education regarding the discipline of

work*development of humor as a self defense mechanism*a proper administration of leisure and self care

One of the ideas related to stress effects on coagulationis prophylactic administration of aspirin in antiplatelet

dose, may be to be started earlier than indicated for othergroups of population.

But the most important lesson of this interesting aspectof our activity is the creation of a proper attitude towardsprofessional stress. Stress seems to be part of our dailyprofessional activity. This is why its management shouldbe an integral part of our education and training.

As the ASA Newsletter specified (2001;65: 13), thetime came for us to reverse the burden of proof in theinterest of safety. The next life we save might be our own.

11. SALIVARY CORTISOL AS A MEASURE OF PROFESSIONALSTRESS; AN OVERVIEW AND A DESCRIPTION OF A STUDYWITH PARAMEDICS

Eliahu Heldman – PhD

Department of Clinical Biochemistry, Ben Gurion University of the

Negev, Beer Sheva, Israel

Work-related stress is caused by an excessive load on theworker and generally is affected by the environment atwork. Work stress has been implicated as a factor incardiovascular diseases, reduced immune functions, met-abolic diseases and mental illnesses. To minimize the risksof such illnesses, particularly those that are work-related,an independent assessment of the stress level of the workeris needed. Since the assessment of the stress level, when aprofessional stress is in question, should be done in theenvironment of the work place, questionnaires and psy-chological evaluations pose difficulties and are essentiallyimpractical with respect to a variety of professions. Forthis reason, biomarkers are needed for the objective,reliable and unproblematic measurement of stress. Cur-rently, cortisol level is the most promising biomarker toassess the response to chronic stress. It is widely acceptedthat stress is associated with changes in circulating cortisollevels, whereas the hormone is quickly elevated imme-diately after the subject is exposed to the stressfulconditions. There seem to be a correlation between thelevels of the stress and the levels of blood cortisol thatfollow the stressful event. Therefore, determination ofcortisol levels may provide a reliable index for the stresslevel of professionals who are exposed to stressful condi-tions during their work. A good correlation has beenshown to exist between the level of the blood cortisol andsalivary cortisol. Since salivary cortisol may be measuredin a non-invasive manner in samples of human saliva thatcan be collected while the worker performs his duty, itmay be an attractive method for the assessment of stress inthe environment of the work place. To test whetherchanges in cortisol levels are indeed correlated with pro-fessional stress, we carried out a pilot study in paramedics,

14 Journal of Clinical Monitoring and Computing

measuring salivary cortisol under conditions of rest andwhile being on duty during an emergency call. The resultsof this study, as well as an overview about the relationshipsbetween salivary cortisol and professional stress, will bediscussed in this presentation.

12. IMPLEMENTATION OF A STANDARDS-BASED,CDA-COMPLIANT ANESTHESIA RECORD

Hurrell M. J.1, Monk T. G.2, Nicol, A.3,

Norton A. N.4, Reich D. L.5, Walsh J. L.6

1Informatics CIS Ltd., Glasgow, UK; 2Duke University Medical

Center, Durham, NC, USA; 3Informatics CIS Ltd., Glasgow,

UK; 4Pilgrim Hospital, Boston, UK; 5Mount Sinai School of

Medicine, New York, NY, USA; 6Massachussets General Hos-

pital, Boston, MA, USA

Introduction: With the increasing use of anaesthesia infor-mation management systems (AIMS) there is the oppor-tunity for different institutions to aggregate and shareinformation both nationally and internationally. Potentialuses of such aggregated data include outcomes research,benchmarking and improvement in clinical practice andpatient safety. However, these goals can only be achievedif data contained in records from different sources are trulycomparable and there is semantic inter-operability. Thispaper describes the development of a standard terminol-ogy for anesthesia and also a domain analysis model andimplementation guide to facilitate a standard representa-tion of AIMS records as XML documents that are com-pliant with the HL7 V3 Clinical Document Architecture(CDA) schema. A representation of vital signs that iscompliant with the ISO 11073 standard and convergencewith the IHE Technical Framework are also discussed.

Methods: In 2002, the Anesthesia Patient Safety Foun-dation (APSF) established the Data Dictionary Task Force(DDTF), to develop a standardized anesthesia terminologyfor use in AIMS. An agreement between APSF andSNOMED International to enhance the anesthesia con-tent of SNOMED CT was signed in 2003. In order toreflect its international membership the DDTF adoptedthe name International Organization for Terminologies inAnesthesia (IOTA). The group’s modus operandi was todevelop and review top-level headings for the terminol-ogy and then to create subsidiary hierarchies of terms ofrelevance to anesthesia. In all cases, the first stage was toreview SNOMED CT content to identify any existingterms that satisfied requirements. Where such matcheswere found the SNOMED concept description andconcept identifier were copied to the IOTA terminol-ogy. A terminology editor was built on the Protegeontology editor using the Protege-OWL plugin and a

custom plugin (developed by the Department ofComputer Science, Manchester University UK). Thelatter provided a user interface that was optimized foruse by clinicians and provided a convenient means tocross-reference terms taken from other classifications.Where terms did not exist in SNOMED CT thesewere added to the IOTA terminology as native IOTAterms and subsequently submitted to the SNOMED CTExtensions Board for consideration for inclusion infuture releases.

In parallel with this activity the HL7 working group forthe Generation of Anesthesia Standards (GAS WG)developed a set of anesthetic use cases and a domaininformation model and also defined structural aspects ofthe anesthetic record. The GAS WG worked closely withthe HL7 Healthcare Devices WG to define ways in whichthe ISO 11073 Terminology and Domain InformationModel (DIM) could be applied to the representation ofvital signs in the anesthetic record. In 2009 the HL7Technical Steering Committee approved a formal projectto develop and publish a CDA Implementation Guide toassist those wishing to create HL7 CDA-compliant anes-thetic records.

Results: The IOTA term set now comprises approxi-mately 4,500 terms that it are intended to meet mostrequirements. Nearly all of these terms are mapped toSNOMED CT. Work continues to increase coverage forsome more specialized areas.

A prototype example of an anesthetic record, renderedas a CDA-compliant XML document has been createdthat reflects a general use case (general anesthetic, repair ofaortic aneurysm). This has been used in a pilot project todemonstrate the feasibility of transferring informationfrom an AIMS system to the US National Surgical QualityImprovement Program (NSQIP) database. Work on therepresentation of vital signs data compliant with the HL7RIM/CDA and the ISO 11073 DIM is well advanced.

Discussion: It is anticipated that the first draft of animplementation guide for CDA-compliant anestheticrecords will be available during 2010. This will also definevalue sets for the various elements of the record takenfrom the IOTA terminology. It is hoped that this willfacilitate the sharing and aggregation of anesthetic datawith consequent improvements in the quality of patientcare.

REFERENCES

1. Gardner M., Peachey T. A Standard XML Schema forcomputerised anaesthetic records. Anaesthesia, 2002,57, pp 1174–1182

Abstracts 15

2. Walsh JL, Hurrell MJ, Wu H, Tomeh M, Monk T.Mapping Anesthesia Records to Standard Terms En-ables Inclusion of AIMS Data in the NSQIP Database.Abstract presented at the American Society of Anes-thesiologists Annual Meeting, 2009. New Orleans, LA.

13. ANALGESIA NOCICEPTION INDEX ONLINE COMPUTATIONAND PRELIMINARY CLINICAL TEST DURINGCHOLECYSTECTOMY UNDER REMIFENTANIL-PROPOFOLANAESTHESIA

Jeanne, M.1,2, Clement, C.1, De Jonckheere, J.2,

Logier, R.2, Tavernier B.1

1Pole d’Anesthesie Reanimation Roger Salengro, University Hos-

pital, Lille, France; 2Inserm CIC-IT 807 University Hospital,

Lille, France

Introduction: During general anaesthesia (GA), the reac-tions of the autonomous nervous system (ANS) to noci-ceptive stress are dampened by opioid administration, butopioid overdose can impair hemodynamic status andmight induce post operative hyperalgesia. The search forthe minimal efficient dose of opioid has become anobjective of modern anaesthesia. Up to now, there is novalidated, routinely used analgesia monitor. Clinicalparameters such as Heart Rate (HR) and Arterial BloodPressure (ABP) are used to assess the need for opioid, butneither are sensitive or specific of pain and analgesia.Several studies have shown that Heart Rate Variability(HRV) analysis gives information related to the ANSactivity [1–2]. The strong influence of anaesthetic drugsover the ANS led some authors to test whether HRVcould be used as an anaesthesia global depth measure[3–4], but to our knowledge, none has described a mea-sure of the pain/analgesia balance during GA. HRV ismediated primarily by changing levels of parasympatheticand sympathetic outflow from the central nervous systemto the sinoatrial node of the heart. In adults, growingevidence highlights that a high frequency power decrease(0.15 – 0.5 Hz) during unpleasant stimuli or emotions isrelated with a drop of vagal tone [5]. During surgicalprocedures in adult patients, HRV is correlated with thebalance between the nociceptive stimulus and the level ofanalgesia [6]. We have previously described and evaluateda pain/analgesia balance measurement algorithm usingHRV analysis [7].

In this paper, we present the development of a newmonitoring system (PhysioDoloris) based on the previ-ously described technology, giving an original AnalgesiaNociception Index (ANI) for nociception/analgesia bal-ance online assessment during GA.

MethodECG acquisition and ANI computation: the whole online

ANI computation process relies on a real time ECGanalysis. The 250 Hz digitized ECG is used for R wavedetection, automatic filtering [8] and computation of theRR series. The raw RR series is mean centred andresampled at 8 Hz, normalised and then band pass filteredbetween [0.15–0.50 Hz]. When parasympathetic tone ispresent, each respiratory cycle influences the RRHF seriesand causes a brief decrease in heart period (figure 1, upperpanel). In case of parasympathetic tone decrease, theinfluence of each respiratory cycle is dampened (Figure 1,lower panel) [7].

The relative parasympathetic tone is assessed by com-puting the area under the normalised RRHF series curveas shown in Figure 1: local minima and maxima are de-tected, and the areas A1, A2, A3 and A4 are measured asthe area between the lower and upper envelopes in each16 s sub-window. We defined AUCmin = min(A1,A2,A3,A4). ANI is then computed in order to expressthe fraction of the total window surface occupied byrespiratory influenced RRHF series, which leads to ameasure between 0 and 100 : ANI = 100 * [a*AUC-min + b]/12.8

The constant a = 5.1 and b = 1.2 have been deter-mined in order to keep coherent the qualitative visualeffect of respiratory influence on RRHF series and thequantitative measurement of ANI.

Preliminary clinical test: after institutional approval, aprototype has been evaluated in a prospective, noninterventional clinical trial on patients planned to undergogeneral anaesthesia for urgent laparoscopic cholecystec-tomy. All patients gave written informed consent.Anaesthetic protocol comprised propofol and remifentanildelivered by a target controlled device (Orchestra� BasePrimea, Fresenius Kabi, France). The surgical and anaes-thetic procedures were not altered by inclusion in thestudy; medical staff was blind to the monitor. Afterinduction of GA, myorelaxation and tracheal intubation,the patient was ventilated in a volume controlled mode.Propofol target was lowered to 2 lg/ml-1 and afterwardsadapted in order to maintain the Bispectral index in thepredefined range of [40–60]. Remifentanil target waslowered at 2 ng.ml-1 and afterwards adapted in case ofhemodynamic reactivity (HemodReact) defined as a 20%increase in heart rate or systolic blood pressure. ANI,hemodynamic and anaesthetic data were recorded duringanaesthesia at predefined time points : (1) after inductionof anaesthesia with volume controlled ventilation; (2)during surgery; (3) in case of HemodReact; (4) after endof surgery but still under anaesthesia with volume con-trolled ventilation. Friedman test for multiple comparisonsand Wilcoxon test for paired comparisons were used. A p

16 Journal of Clinical Monitoring and Computing

value of less than 0.0125 was considered significant.Results are presented as median (interquartile).

Results: nine patients were included, aged 32(7) years.Eight patients presented a HemodReact period duringanaesthesia. Bispectral index remained stable in the pre-defined range at 43(10). Systolic blood pressure was sig-nificantly higher at HemodReact time points as expectedfrom the study design. ANI was significantly lower duringHemodReact time points (Table 1), but also duringsurgery (i.e. before HemodReact) as compared withmeasurements after induction and after end of surgery.Propofol and remifentanil targets did not vary significantlyduring these predefined time points.

Discussion: these preliminary clinical results show thatANI seems to be related with the surgical context of thepatient: parasympathetic tone was highest when nocicep-tion was absent as recorded after induction (1) and after endof surgery (4). Parasympathetic tone decreased whennociception increased during surgery (2), and furtherdecreased (although not significantly) in case of Hemod-React (3). A greater number of inclusions is needed tomeasure the predictive positive values of an ANI decrease.Limitations to ANI measurement are essentially non sinusrythms, non ventilated periods such as during intubationand electric noise produced by the electric knife.

REFERENCES

1. Heart rate variability. Standards of measurement, physio-logical interpretation and clinical use. Circulation, 1996.93(5): 1043–65.

2. Pichot, V., Gaspoz, J.M., Molliex, S., Antoniadis, A.,Busso, T., et al., Wavelet transform to quantify heart ratevariability and to assess its instantaneous changes. J ApplPhysiol, 1999. 86(3): 1081–91.

3. Latson, T.W. and O’Flaherty, D., Effects of surgicalstimulation on autonomic reflex function: assessment ofchanges in heart rate variability. Br J Anaesth, 1993. 70(3):301–5.

4. Pichot, V., Buffiere, S., Gaspoz, J.M., Costes, F.,Molliex, S., et al., Wavelet transform of heart rate vari-ability to assess autonomous system activity does not predictarousal from general anesthesia. Can J Anaesthesia, 2001.48(9): 859–63.

5. Appelhans, B.M. and Luecken, L.J., Heart rate variabilityand pain: associations of two interrelated homeostatic pro-cesses. Biol Psychol, 2008. 77(2): 174–82.

6. Jeanne, M., Logier, R., De Jonckheere, J. andTavernier, B., Heart rate variability during total intravenousanesthesia: effects of nociception and analgesia. AutonNeurosci, 2009. 147(1–2): 91–6.

7. Jeanne, M., Logier, R., De Jonckheere, J. andTavernier, B., Validation of a graphic measurement of heartrate variability to assess analgesia/nociception balance duringgeneral anesthesia. Conf Proc IEEE Eng Med Biol Soc,2009. 1: 1840–3.

8. Logier, R., Dejonckheere, J. and Dassonneville, A. Anefficient algorithm fo R-R interval series filtering. in Pro-ceedings of the 26th Annual International Conference of theIEEE Engineering in Medecine and Biology Society. 2004.

14. PHARMACODYNAMIC MODELING OF CHANGES IN PULSEWAVEFORM DURING INDUCTION OF PROPOFOLANAESTHESIA IN VOLUNTEERS: COMPARISON BETWEENINVASIVE AND CONTINUOUS NON-INVASIVEMEASUREMENTS OF PULSE PRESSURE

Christian Jeleazcov1,2, Andreas Tobola2, Michael

Weiss1, Christian Weigand2, Jurgen Schuettler1

1Department of Anaesthesiology, University Hospital Erlangen,

Germany; 2Fraunhofer Institute for Integrated Circuits and

METEAN, Erlangen, Germany

Introduction: Propofol anaesthesia causes changes in pulsepressure waveform. Features of pulse pressure waveform

Fig. 1 Normalized and filtered RR series in two different states of anal-gesia/nociception balance during general anaesthesia. A1, A2, A3 and A4are the areas measuring the respiratory influence in the RR series; upperpanel: adequate analgesia; lower panel: light analgesia leading to an increaseof HR and ABP.

Table 1 Data presented as median (interquartile)

Fried-man

test

(1) After

induction

(2) During

surgery

(3) Hemod

react

(4) After end

of surgery

HR (bpm) NS 65(9) 65(16) 72(7) 58(21)

SBP (mmHg) <0.01 95(20) 113(24) 130(8)* 124(25)

BIS NS 34(12) 31(11) 41(10) 49(16)

ANI (%) <0.01 88(17) 50(9)*,+ 40(13)*,+ 97(22)

prop Ce (lg ml-1) NS 3.0(0.6) 3.0(0.9) 3.0(0.5) 2.2(1.1)

remi Ce (ng ml-1) NS 4.0(2.1) 3.6(0.7) 4.3(1.1) 3.0(0.8)

* p < 0.01 vs. after induction(Wilcoxon test) + p < 0.01 vs. after end of sur-

gery(Wilcoxon test).

Abstracts 17

from radial artery are associated with blood flow proper-ties of central arteries and left ventricular heart function.[1, 2] Since maintenance of stable hemodynamics is one ofthe main goals of adequate anaesthesia, non-invasivemonitoring of pulse waveform parameters may contribute

additional information to routine measurement of arterialblood pressure.

Continuous non-invasive finger blood pressurerecordings performed with Task Force Monitor� (TFM)provide real-time pulse pressure waveforms. In thisexploratory study, we estimated propofol induced changesin three pulse waveform parameters during induction ofanaesthesia by pharmacokinetic/-dynamic (Pk/Pd) mod-eling and examined whether Pd parameters fitted withnon-invasive measurements were similar to those fittedwith invasive measurements from radial artery.

Fig. 2 Time course of propotol Ce and dPdTmax of ALINE and TFM inone volunteer.

Fig. 1 Investigated pulse waveform parameters.

Fig. 3 .

Table 1 Propofol induced changes on absolute values. * P < 0.05to baseline

Table 2 Goodness of the pharmacodynamic fit

Table 3 Pharmacodynamic parameters ke0, EC50, and gamma ofsigmoid relationship propofol concentration—pulse waveform para-meters PP, dPdTmax, and AUC diast

18 Journal of Clinical Monitoring and Computing

Methods: After local ethics committee approval westudied 9 volunteers of ASA I. Propofol was infusedvia target controlled infusion to achieve plasma concen-trations increasing from 0.5 to 4.5 lg/ml in steps of0.5 lg/ml. Each infusion step was maintained constant forat least 15 min. Following the last step, propofol wasfurther linearly increased by 0.5 lg/mlÆmin, until one ofthe following conditions was present: EEG suppres-sion > 2 s, flattening of spontaneous breathing, or meanblood pressure drop > 45% from baseline.

Pulse pressure (PP), average change in pulse pressure(dPdTmax), and area under diastolic curve (AUCdiast)(Figure 1) were derived from valid pulse pressure beatsthat were recorded from continuous invasive radial(ALINE) and non-invasive TFM measurements of sameupper extremity. The relation between drug dosing andtime course of pulse waveform parameters was modeledusing a Pk/Pd model with a common sigmoid concen-tration-effect relationship in which EC50 is the concen-tration for half-maximum effect, and c (gamma) exponentdescribes the steepness of the concentration effect. Weexpressed the propofol effect on pulse waveform param-eter (P) in each volunteer as the percent decrease frombaseline: P(t)/Pbaseline 9 100 = 100 – [100 9 Ce(t)

c]/EC50

c + Ce(t)c]. Ce(t) is the effect site concentration

obtained by differential equation dCe/dt = (Cpl - Ce) Æke0, whereby Cpl is measured plasma concentration ofpropofol, and ke0 denotes the first-order rate constantdetermining the efflux from the effect-site. [3] Thegoodness of the pharmacodynamic fit was assessed bythe unweighted residuals R = (EM-EP)/E0 9 100% andthe absolute residuals AR = (|EM- EP|)/E0 9 100%from E0, EM, and EP as baseline, measured andpredicted effect, respectively. Significant differencesbetween ALINE and TFM were identified by Wilco-xon signed-rank test. The fitting procedure and statis-tical calculations at a significance level of a = 0.05 havebeen performed with Matlab (Version 2009a, TheMathworks Inc., Natick, MA, USA). Data are presentedas mean ± SD (median).

Results: Propofol induced a significant decrease in PP,dPdTmax, and AUCdiast (Table. 1, Figure 2). Figure 3depicts the sigmoid relationship between propofol con-centration and one of the Pd parameters (dPdTmax) inone volunteer. The results of Pd modeling are summa-rized in Tables 2 and 3. The observed weak differences inke0, EC50 of dPdTmax between ALINE and TFM failedto achieve statistical significance (p = 0.07).

Conclusion: Continuous non-invasive monitoring ofpulse waveform parameters may be useful for pharmaco-dynamic studies on drug induced changes of pulsewaveform and may add useful information to routinemeasurement of arterial blood pressure.

REFERENCES

1. McDonald’s Blood Flow in Arteries: Theoretical,Experimental and Clinical Principles. 5th edition.2005, Hodder Arnold: London.

2. Hashimoto, J., Y. Imai, and M.F. O’Rourke, Indicesof pulse wave analysis are better predictors of leftventricular mass reduction than cuff pressure. Am JHypertens, 2007. 20(4): p. 378–84.

3. Jeleazcov, C., et al. Pharmacodynamic modelling ofchanges in arterial blood pressure during propofolanesthesia in volunteers: comparison between invasiveand continuous noninvasive measurements. WC 2009,IFMBE Proceedings 25/VII: 582–5

15. ASSESSING PERFORMANCES OF GLUCOSE CONTROLALGORITHMS ON A SET OF VIRTUAL ICU PATIENTS

A. Guerrini1,2, M. Sorine1, P. Kalfon3

I.N.R.I.A. Paris-Rocquencourt, France; 2LK2, Saint-Avertin,

France; 3 CH Louis Pasteur, Chartres, France

Introduction: The 2001 van den Berghe et al. study showedthat intensive glucose control reduced mortality. Sub-sequent large randomized controlled trials of tight glucosecontrol have failed to show a mortality benefit. Conse-quently, considerable controversy has emerged as towhether tight glucose control is warranted in all criticallyill adults. Differences between studies have been explainedby the expertness of medical staff. Therefore, with nonexpert nurses, glucose control is solely based on algorithmperformance. Rather than an educational simulationprogram of insulin dosage and dietary adjustment (AIDA[1]), we provided physicians with a tool to score algorithmperformance using realistic virtual ICU patients beforeactual trials. Our simulation was conducted using tworecent algorithms as examples: the NICE-SUGAR algo-rithm [2] and the one used in the ongoing CGAO-REAstudy [3], which has been developed using our tool. Ourmain objective was to add a validation step ahead of trialsin a real environment setting by verifying the algorithmmanages to control virtual ICU patients.

Methods: The set of virtual patients was built using realdata coming from Chartres Hospital patients using a non-linear pharmaco-dynamic glucose-insulin system model [4]where patients’ endogenous glucose clearance and insulin-sensitivity were time dependent parameters. Algorithmrobustness was then tested on ten diabetic and non-diabeticvirtual patients by simulating the ICU environment: 1)delays were implemented by generating random numbersusing the delays distribution law observed in Chartres

Abstracts 19

Hospital, while 2) inaccuracy of glucose meters was used asperturbation. Then, recommendations computed by eachtested algorithm were applied to the virtual patients and theoverall performance over the whole stay was assessedaccording to standard scores (e.g. time spent in the targetrange) and indexes (e.g. variability index).

Results: For highly insulin-sensitive patients or patientswith high variation of insulin sensitivity, scores obtainedusing NICE-SUGAR algorithm were very low. The per-centage of time in the target range (4.4–6.1 mmol/L) waslower than 50% for almost all patients and the frequency ofhypoglycemia was high. With insulin-sensitive patients,glucose fluctuations and sometimes severe hypoglycemia(see Figure 1) were induced by the algorithm. The meanpercentage of time in the target range using CGAO systemwas about 60% and the variability index calculated in thiscase was significantly lower than with NICE-SUGAR.NICE-SUGAR control showed random behavior consid-ering delays and noise and no trend was highlighted whereasCGAO system was robust to noise and delays in all cases.

Discussion: NICE-SUGAR outcome has cast doubt onwhether tight glucose control is beneficial and a new targetrange has been suggested subsequent to its publication.However, control algorithms are at the forefront of bloodglucose target achievement problems and thus, a targetrange without a proper algorithm to achieve it is useless.The purpose of this study is to demonstrate that preliminarytrials, such as numerical testing, enable the invalidation oflow-scored algorithms and consequently, avoid targetoverlay between groups. Numerical simulations of theNICE-SUGAR algorithm agreed with the post-analysis byrevealing relatively high glucose variability outside thetarget range in some cases, which increases the risk ofhypoglycemia. In contrast, scores obtained using theCGAO-REA algorithm were encouraging and further tests

were performed on real ICU patients prior to the start ofthe CGAO-REA study. Finally, although the effect of nonrespect of insulin flow recommendations have not beeninvestigated, it is likely that this factor played an importantrole in glucose control.

REFERENCES

1. AIDA Diabetic software simulator program of bloodglucose-insulin interaction (http: //www.2aida.net)

2. Normoglycaemia in Intensive Care Evaluation andSurvival Using Glucose Algorithm Regulation - NICE-SUGAR (https: //studies.thegeorgeinstitute.org/nice)

3. Computerized Glucose Control in Critically Ill Patients(CGAO-REA) (http: //www.clinicaltrials.gov/ct2/show/NCT01002482)

4. C.E. Hann, J.G. Chase, J. Lin, T. Lotz, C.V. Doranand G.M. Shaw, Integral-based parameter identifica-tion for long-term dynamic verification of a glucose–insulin system model, Comput. Methods ProgramsBiomed. 77 (3) (2005), pp. 259–270.

16. AUTOMATION AND AUTOMATION SURPRISES: LESSONSFROM AVIATION: SHOULD HEALTH CARE BRACE ITSELF?

Cor J. Kalkman

Perioperative & Emergency Medicine, University Medical Centre

Utrecht, Utrecht, The Netherlands; c.j.kalkman@umcutrecht.nl

The immense growth in computing power and software toolshas virtually eliminated ‘steam gauges’ in the cockpits ofmodern aircraft and it is rapidly encroaching upon health care.This means that controls, dials and levers are no longer con-nected to mechanical components of medical devices, but arejust control inputs for software that controls the device. This isimmediately apparent in the last generation of ventilators thathave discarded traditional rotameters to indicate gas flows infavour of software controlled flow control valves and on-screen graphical representations of gas flow, often using bargraphs to mimick the familiar mental model of a rotameter.Rather than setting flow rates individually, modern ventilatorsallow the user to set inspired oxygen concentrations or an‘economically optimized’ gas flow, while the software cal-culates and controls actual gas flows.

In contrast, computerized drug delivery and closed-loopcontrol solutions have lagged behind, mainly as a result offear for adverse events and possible litigation. There is alsomore debate as to whether such closed-loop control systemsare solutions for actual clinical problems or just ‘gadgets’ for

Figure 1.

20 Journal of Clinical Monitoring and Computing

technically talented anaesthesiologists. Closed-loop control ofneuromuscular blockade is certainly feasible1,2 and relativelysafe, although accidental overdosing will necessitate postop-erative mechanical ventilation. In contrast, errors with closed-loop administration of vasoactive drugs could kill a patient.Potential sources of error are unreliable blood pressure data orundetected software errors. Interestingly, even relativelysimple model-based ‘target- controlled’ infusion of hypnoticsthat has long been available to anaesthesiologists in Europe, isstill not approved in the US by the Food and Drug Admin-istration. Much to the chagrin of TCI enthusiasts, FDAemployees have expressed vaguely defined fears such as‘‘important health implications’’, ‘‘significant incrementalrisk’’ of anaesthetic controllers, ‘‘combining high level lan-guages, general purpose computers, and complex operatingsystems that might result in products that are too elaborate forthe developer to verify entirely’’3. The logical result is thatmanufacturers have been discouraged from developing im-proved systems and entering this market and closed-loopcontrol of hypnotics remains in the experimental domain.

Automation versus the experienced human operator: who is in charge?Several notorious aircraft disasters were clearly related

to problems at the interface between the human operatorsand their automated flight management systems. Ameri-can Airlines flight 965 crashed into a mountain near Cali,Colombia because the crew had first entered the wrongwaypoint code for the non directional beacon ‘Rozo’ inthe FMS (‘R’ – as indicated on their charts - instead of‘ROZO’- the most current identifier) which took theminto the direction of Bogota. When the crew discoveredthe error, they entered the coordinates of the Cali airportin the FMS; the aircraft changed course to Cali, but theaircraft was now to the east of mountain ridge and theplane crashed into a mountain with the loss of 189 lives.

Gol Transportes Aereos Flight 1907, a Boeing 737 col-lided in mid-air with a business jet over the Brazilianrainforest. All 154 passengers and crew were killed. Inves-tigation of this crash revealed that the crew of the brand newEmbraer business jet had failed to notice a tiny text messageannouncing that the automatic Traffic Collision AvoidanceSystem was inoperative. Both planes had difficulty com-municating with air traffic control. As a result, both planes -although flying in opposite directions - were on the samealtitude, which should not have happened. As a result of‘perfect’ flight guidance by GPS and FMS automation, bothplanes were flying exact opposite headings at exactly thesame altitude.

More recently, a Boeing 737 from Turkish Airlinescrashed 1.5 km north of runway 18R at Schiphol airport,the Netherlands, because a defective radio altimeter - reg-istering minus 8 feet altitude and thus assuming that theaircraft had landed on the runway - repeatedly closed theautothrottle causing the aircraft to sink below the glides-

lope. The crew reacted too late to this automation surprise,the aircraft stalled and crashed into a field north of therunway threshold killing 9 people including the captain andfirst officer. The accident investigation also blamed thecrew for reacting too late to the failing automation.

Automation: who is in charge? Who is responsible?There seems to be a dividing line between Europe and

the US regarding the amount of control we are willing tohand over to automation. The Americans seem to be moredistrustful of computers taking control than Europeans.Nonetheless, it is impossible today to make any scheduledflight in a modern jet aircraft without entrusting one’s lifeto the proper functioning of a large amount of computersystems, both in the plane and on the ground. The twomajor aircraft manufacturers, Boeing and Airbus, haveopposing views as to the degree to which the flight datamanagement computers should be able to overrule a pilotwhen he is about to subject the aircraft to dangerousextreme attitudes and control movements. While eachaircraft will generate alerts and voice alarms (‘‘pull up!’’,‘‘too low, terrain!’’), the automation of a Boeing will nevertake control away from the pilot and will grant him thepower to decide how to steer the plane. In contrast, Airbushas decided that the safety of the flight is best guaranteed ifthe automation can overrule a pilot who is about to per-form a very dangerous manoeuvre. These differences in‘control philosophy’ will soon become important in healthcare. Designers of anaesthesia machines and delivery sys-tems are already confronted with such decisions. Similardesign issues also confront software developers who buildhospital-wide electronic patient record systems. Should adoctor be blocked from prescribing penicillin to a patientwhen the ‘allergy’ data field contains: ‘allergic to penicil-lin’, or should she only be warned and required to elec-tronically sign a motivated waiver?

Taken together these issues constitute the field of HumanFactors. It is remarkable how many industries have em-braced the science of Human Factors to help them improveman–machine interaction, the design of user interfaces andteam interaction. Health care has been slow in adoptingHuman Factors knowledge into the design of its processesand equipment. The rapid rise in complexity of care, newtechnology and increasing pressures to produce both effi-cient and safe care, make it necessary to better reflect on thedesign of interfaces to ICT systems and medical devices.

Unintended consequences of technology: problems and pitfallsof automation

As we rapidly move towards hospital wide electronicpatient records (EPR) and Computerized PhysicianOrder entry (CPOE), new types of errors can occur.Several studies have shown that CPOE reduces thenumber of medication error, but it is unclear to whatextent serious new errors resulting from working with a

Abstracts 21

computer interface might counterbalance the benefits ofCPOE. To select a drug, so-called ‘pick lists’ are used inpull-down menu’s. It is relatively easy to accidentallyselect the item below or above the correct item in analphabetic pick list, especially in the case of ‘sound alike,look alike’ drugs. A similar problem might occur withpreformatted order sets.

Will automation promote complacency? When automatedblood pressure monitors appeared, taking a blood pressurewas no longer an act that involved feeling the patient’s pulse.Some experienced anaesthesiologists expressed fear thatautomated blood pressure monitoring would reduce vigi-lance and lead to anaesthesia practice that is less ‘hands-on’.Similar objections have been voiced against automatedanaesthesia record keeping systems or closed-loop control ofanesthesia administration: they might discourage systematicscanning of the monitors and promote complacency. Fewstudies have addressed this issue. Loeb et al. found no evi-dence for reduced vigilance when anaesthesia residents wererelieved of the task of charting the anaesthesia record.4.Weinger et al. also were unable to document a reduction invigilance when an ARK was used during anaesthesia forcardiac surgery5

What if the system fails: are we ready?The more hospitals start to rely on ICT solutions for their

clinical processes, the more critical such systems become forsafe patient care. Beth Israel Hospital in Boston, MA was anearly adopter of electronic patient records and its residents hadnever used paper-based records or ordering. In November2002 the network was down for 4 days, forcing doctors andnurses to revert back to pen and paper. The cause of theproblem was later discovered. A researcher had started toupload several gigabytes of data and the process was stuck in anendless loop blocking the network. Separating databases forresearch and clinical use clearly is mandatory, using ‘mirror’databases for all clinical research queries.

The issue of how to balance the need for redundancywith the additional costs is also unresolved. Pagers andanalog telephone lines are on the brink becoming extinctand new technologies such as voice over IP (VoIP) areready to take over. The risks of entrusting all communi-cations (data and voice) to a single hospital network areimmense. If today a part of the hospital information systemfails, doctors and nurses can still use the telephone tocommunicate with colleagues, the emergency room, ICUor the laboratory. If the network is also used for voicecommunication and the network is down, all communi-cation except talking face to face becomes impossible.

ConclusionsComputers and automation are now an integral part of

anaesthesia and intensive care technology. Computercontrol is already built into many new anaesthesia devices,but closed-loop control of critical anaesthesia processes

such infusion of vaso-active drugs or hypnotics remains inthe experimental domain. Computerized patient datamanagement systems are rapidly becoming standard ofcare and will increasingly incorporate decision supportalgorithms in the form of reminders and alerts. Suchsystems offer a promise of considerable clinical benefit,but we do not know yet the risk–benefit ratio of allowingICT to overrule the doctor. This wave of ICT also createsnew challenges, as we discover the unintended conse-quences of this technology. Unique new types of errorsuch as selecting the wrong drug from an alphabetic picklist, will require new solutions. Finally, as hospitals be-come dependent on the correct functioning of theirnetwork and tightly coupled ICT subsystems, sufficientredundancy needs to built into the system.

REFERENCES

1. Ebeling BJ, Muller W, Tonner P, Olkkola KT, StoeckelH: Adaptive feedback-controlled infusion versusrepetitive injections of vecuronium in patients duringisoflurane anesthesia. J Clin Anesth 1991; 3: 181–5

2. Lendl M, Schwarz UH, Romeiser HJ, Unbehauen R,Georgieff M, Geldner GF: Nonlinear model-basedpredictive control of non-depolarizing muscle relax-ants using neural networks. J Clin Monit Comput1999; 15: 271–8

3. Egan TD, Shafer SL: Target-controlled infusions forintravenous anesthetics: surfing USA not! Anesthesi-ology 2003; 99: 1039–41

4. Loeb RG: A measure of intraoperative attention tomonitor displays. Anesth Analg 1993; 76: 337–41

5. Weinger MB, Herndon OW, Gaba DM: The effect ofelectronic record keeping and transesophageal echo-cardiography on task distribution, workload, and vig-ilance during cardiac anesthesia. Anesthesiology 1997;87: 144–55; discussion 29A–30A

17. USE OF THE INVENT SYSTEM FOR STANDARDIZEDQUANTIFICATION OF CLINICAL PREFERENCES TOWARDSMECHANICAL VENTILATOR SETTINGS

C. Allerød1,2, D. S. Karbing1, P. Thorgaard2,

S. Andreassen1, S. Kjærgaard2, S. E. Rees1

1Center for Model-based Medical Decision Support, Department of

Health Science and Technology, Aalborg University, Aalborg,

Denmark; 2Department of Anaesthesia, Aalborg Hospital, Aarhus

University, Aalborg, Denmark

Introduction: Selecting appropriate mechanical ventilatorsettings is a difficult task requiring a compromise between

22 Journal of Clinical Monitoring and Computing

conflicting clinical goals of securing gas exchange andavoiding ventilator induced lung injury. Guidelines havebeen successfully evaluated for acute lung injury and acuterespiratory distress syndrome [1]. Observational studieshave shown these guidelines are slow to be fully acceptedand employed in clinical practice [e.g. 2], but these studiesare unable to compare different clinicians’ preferencesunder identical circumstances. The purpose of this workwas to build a computer system for standardized quanti-fication of clinical preferences towards mechanical venti-lator settings. The system is based on the INVENTdecision support system [3], which includes physiologicalmodels enabling simulation of patients’ response to chan-ges in ventilator settings. The system was used to quantifyclinicians’ preferences for inspired oxygen fraction (FiO2),tidal volume (Vt) and respiratory frequency (f), whichwere then compared to suggestions by INVENT.

Methods: Figure 1 illustrates the use of physiologicalmodels for simulating mechanically ventilated patients.Three models are used describing pulmonary gasexchange, acid–base chemistry of blood and lungmechanics. Clinicians can get an overview of the patientstate described by model parameters and measured val-ues. The system simulates changes in patient state forvariations in FiO2, Vt and f. Clinicians can vary thethree settings until the preferred combination of settingsand simulated outcome are found. Preferred settings areautomatically registered for later analysis. The systemrequires clinicians to assume model simulations are cor-rect, accept levels of PEEP and I: E ratio as appropriateand assume patients weigh 70 kg.

Results: The system was successfully used in a studyquantifying preferences of 10 senior intensive care cli-nicians from the 4 university hospitals in Denmark. The

clinicians individually selected their preferred settings in10 real patient cases described by the models and pre-sented to the clinicians in a random order. Afterwardseach clinician was presented with the 9 other clinicians’and INVENT’s selected ventilator settings and resultingoutcomes, and were requested to rank the selections.The registered ventilator settings selected by cliniciansand INVENT varied substantially and the subsequentranking (1–11) by clinicians for each patient caseshowed a large variability in what was considered bestpractice. INVENT had the 3rd best average rank withrankings ranging from 3 to 10.

Discussion: The results indicate a lack of consensus onwhat is considered the best levels of FiO2, Vt and f. Thestudy omits selection of PEEP and I: E ratio, but includingthese settings would unlikely lead to greater consensus.The study furthermore demonstrates the possible use ofphysiological models for standardized evaluation of clini-cal preferences and as a tool to reach consensus. Themathematical functions describing clinical preferences in adecision support system as INVENT could then be tunedto fit this new consensus.

REFERENCES

1. The ARDS Network: Ventilation with lower tidalvolumes as compared with traditional tidal volumes foracute lung injury and the acute respiratory distresssyndrome. N Engl J Med 2000; 342: 1301–8.

2. Esteban A et al. Evolution of mechanical ventilation inresponse to clinical research. Am J Respir Crit CareMed 2008; 177: 170–7.

Fig. 1 The use of physiological models for simulating mechanically ventilated patients. Finding the most appropriate settings is an iterative process asillustrated by the dashed arrow.

Abstracts 23

3. Rees SE et al. Using physiological models and decisiontheory for selecting appropriate ventilator settings. JClin Monit Comput 2006; 20: 421–9.

18. MOODS AND BURNOUT AMONG PHYSICIANS:ASSOCIATIONS WITH PRESCRIBING MEDICATIONS,COMMUNICATING WITH PATIENTS, AND REFERRALS FORSPECIALISTS AND DIAGNOSTIC TESTS

Talma Kushnir

Faculty of Health Sciences, Ben-Gurion University of the Negev,

Beer-Sheva, Israel

Introduction: The quality of health care can be compromisedby many factors and is a universal concern. Feelings andemotions may affect behavior in all medical encounters,although many health-care givers still believe that profes-sional conduct is detached from subjective influences and isdetermined mainly by objective considerations. Indeed, thecontribution of affective factors to physician behavior inmedical settings has been studied empirically only infre-quently. Affects include both transitory emotions, such asmoods; and more stable states, such as burnout.

Burnout prevalence has been studied extensively in themedical literature. Physicians in all specialities were found tobe vulnerable, as they are exposed to chronic stresses in themedical environment. It is commonly assumed that burnouthas significant negative consequences on physician perfor-mance, but there is very little research concerning thisassumption using objective outcome measures in healthcaresettings. There is even less research on the effects of transitorymoods on physician behavior. In this presentation we willdescribe some findings from two recent studies that includedobjective and subjective data concerning associations be-tween affective conditions and several physician behavior visa vis patients: referring for diagnostic tests and for specialists,talking, and presecribing medications.

Methods: The first study included 136 primary care physi-cians who responded in an interview to questionnairesassessing burnout and work and personal characteristics. Therates of prescriptions and referrals for diagnostic tests andspecialists were obtained from the HMO data bases. Thesecond study included 188 primary care physicians whoresponded to a self-report questionnaire assessing burnout andperceived behavior under positive and negative moods.

Results: In the first study we found that burnout waspredicted by subjective workload and job satisfaction.Several objectively assessed referral behaviors were asso-ciated with burnout in correlational analyses. In multi-variate analyses, only the rate of referrals to expensiveimaging tests (e.g. MRI) was predicted independently andmodestly by burnout.

In the second study, in five ANOVAs with repeatedmeasures on mood states, the physicians reported that ongood mood compared with negative mood days, theytalked more, prescribed less and referred less (for all bea-hviors, p < .001). High compared with low burnoutphysicians had higher perceived rates of all referralbehaviors. Significant mood*burnout interactions indi-cated that the effects of mood were stronger among highcompared with low burnout physicians.

Discussion and possible implications: In the second studymoods were perceived as having significant but differenteffects on each physician behavior. The negative mooddecreased talking and increased prescribing and referralbehaviors, and vice versa for the positive mood – it in-creased talking ‘‘at the expense’’ of prescribing medica-tions and referring to tests and specialists. Burnoutintensified the effects of moods. Although the literaturefocuses on burnout, transitory moods may have strongereffects on physician behavior and further studies areneeded. Associations between affective states and referralbehaviors should be further studied in larger samples,additional medical specialties and high quality data onreferals and prescriptions.

Altogether, the above findings about the significant effectsof moods, the intensification of mood effects by burnout, andthe association between burnout and referrals to expensiveimaging tests (e.g. MRI) suggest that the incremental effects ofnegative moods and burnout may impair quality of healthcareand may be costly to health services.

REFERENCE

Kushnir, T., Kushnir, J., Sarel, A., & Cohen, H. A.Exploring physician perceptions of the impact of emotions onbehaviour during interactions with patients. Family Practice2010; doi: 10.1093/fampra/cmq070

19. NON-INVASIVE PULSE CONTOUR CARDIAC OUTPUT BYNEXFIN TECHNOLOGY

Johannes J. van Lieshout, MD, PhD1,2, Wilbert A.

Wesselink, MSc, PhD3 & Berend E. Westerhof,

MSc, PhD2,3

1Special Medical Care, Department of Internal Medicine, F7-205,

Academic Medical Center, University of Amsterdam, PO Box

22700, 1100 DE Amsterdam, The Netherlands; tel. (31) 20-

5662371; fax (31) 20-5669223; j.j.vanlieshout@amc.uva.nl;2Laboratory for Clinical Cardiovascular Physiology, AMC Center

for Heart Failure Research, Amsterdam, The Netherlands;3BMEYE bv, Amsterdam, The Netherlands

24 Journal of Clinical Monitoring and Computing

In many patients, continuous knowledge of the hemo-dynamic status would be of great value, however, invasivemeasurement of blood pressure (BP) or cardiac output(CO) may not be warranted. BP can be measured inter-mittently with an upper arm cuff. However, continuousnon-invasive measurement of arterial pressure in humansis available since the early 1980 s when Wesseling et al. [1]introduced the first Finapres device based on the volume-clamp method invented by the Czech physiologist JanPenaz [2]. Monitoring of a continuous CO allows for thedetection of rapid changes in systemic flow that wouldotherwise be unnoticed by the recording of arterialpressure and heart rate [3]. Non-invasive and continuoustracking of changes in stroke volume (SV) can be obtainedby ultrasound, thoracic electrical impedance or by mini-mally invasive methods that use arterial pulse wave anal-ysis from an arterial line that is already in place for BPmonitoring or sampling of blood gasses. Recently a devicebecame available for truly noninvasive and continuousmonitoring of BP and CO, the Nexfin HD monitor [4;5].We will review past and present developments in themethodology of finger pressure and continuous strokevolume/cardiac output by focusing on the methods andtechnologies that the Nexfin uses to measure noninvasive,beat-to-beat hemodynamics.

Nexfin uses the established volume clamp methodologyand Physiocal criteria [6] for measurement of continuousBP in a finger. An inflatable cuff around the finger with anoptical blood volume measuring system clamps the bloodvolume to a preset level by applying a pressure equal toarterial pressure throughout the cardiac cycle. In combi-nation with Physiocal, calibrated recordings of the entirefinger arterial pressure wave are obtained [6]. As a nextlogical step brachial pressure reconstruction was developed,which counteracted the pressure wave amplification inperipheral measurement sites like the finger. In addition,application of a level correction compensates for the pres-sure drop due to resistance to flow in the smaller arteries.The combination of these two methods results in pressuresthat are comparable to invasively measured brachial or radialBP. This resolves issues with earlier devices, which some-times displayed mean finger arterial pressures that weremuch lower than invasive pressures. Further importantimprovements were established by an update of the cuffdesign using modern materials and optical components.

Nexfin uses a recently developed pulse contour methodto determine CO (Nexfin CO-trek), building on earlierdevelopments. The area under the systolic part of thepressure curve is divided by the aortic input impedance toobtain stroke volume, similar to the cZ pulse contourmethod (early Seventies). This cZ method used a simple

formula containing heart rate, mean arterial pressure andpatient age to determine input impedance. Tracking ofchanges was very good, but the absolute values could havea bias. More recent, the Modelflow method (Nineties)used a 3-element Windkessel description of aortic inputimpedance to calculate a flow curve from the pressurecurve [7–10]. Integration of the flow curve yields strokevolume. In the Windkessel, the nonlinear pressure-arearelation of the aorta was described as function of pressure,age, gender, height and weight assuring better absoluteCO values and tracking of changes over wider ranges ofpressure changes. The Windkessel model, incorporatingnonlinear pressure dependency and patient data, is alsoused with the Nexfin CO-trek method, although in thiscase to calculate aortic input impedance. Combined withthe area under the systolic part of the pressure curve strokevolume can now be calculated. Whereas earlier methodswere developed to use invasively measured pressures,CO-trek method operates equally well with noninvasivemeasured BP, enabling routine clinical application [5].

REFERENCES

1. Wesseling, K.H. (1995) A century of noninvasivearterial pressure measurement: from Marey to Penazand Finapres. Homeostasis 36, 2–3.

2. Penaz, J. Photoelectric measurement of blood pressure,volume and flow in the finger. Albert, R., Vogt, W.S., and Helbig, W. Digest of the International Con-ference on Medicine and Biological Engineering, 104.1973. Dresden. 1973.

3. Kim, Y.S., Immink, R.V., Jellema, W.T. andVan Lieshout, J.J. (2007) A perspective on non-inva-sive continuous cardiac output in the critically ill. NethJ Crit Care 11, 185–191.

4. Eeftinck Schattenkerk, D.W., Van Lieshout, J.J.,Van den Meiracker, A.H. et al. (2009) Nexfin nonin-vasive continuous blood pressure validated againstRiva-Rocci/Korotkoff. Am J Hypertens. 22, 378–383.

5. Bogert, L. W., Wesseling, K. H., Schraa, O.,Van Lieshout, E. J., De Mol, B. A., Ven Goudoever, J.,Westerhof, B. E., and Van Lieshout, J. J. Pulse contourcardiac output derived from non-invasive arterialpressure in cardiovascular disease. Anaesthesia. 2010.

6. Bogert,L.W. and Van Lieshout,J.J. (2005) Non-invasivepulsatile arterial pressure and stroke volume changesfrom the human finger. Exp Physiol 90, 437–446.

7. Wesseling, K.H., Jansen, J.R.C., Settels, J.J. andSchreuder, J.J. (1993) Computation of aortic flow

Abstracts 25

from pressure in humans using a nonlinear, three-element model. J Appl Physiol 74, 2566-2573.

8. Jellema, W.T., Groeneveld, A.B., Wesseling, K.H.,Thijs, L.G., Westerhof, N. and Van Lieshout, J.J.(2006) Heterogeneity and prediction of hemodynamicresponses to dobutamine in patients with septic shock.Crit Care Med. 34, 2392–2398.

9. Jellema, W.T., Wesseling, K.H., Groeneveld, A.B.,Stoutenbeek, C.P., Thijs, L.G. and Van Lieshout, J.J.(1999) Continuous cardiac output in septic shock bysimulating a model of the aortic input impedance: acomparison with bolus injection thermodilution.Anesthesiology 90, 1317–1328.

10. Harms, M.P.M., Wesseling, K.H., Pott, F. et al.(1999) Continuous stroke volume monitoring bymodelling flow from non-invasive measurement ofarterial pressure in humans under orthostatic stress.Clin Sci 97, 291–301.

20. MONITORING MITOCHONDRIAL OXYGENATION

Egbert G. Mik

Department of Anesthesiology, Erasmus Medical Center, Rotterdam,

The Netherlands

Introduction: The quantitative assessment of cellularoxygenation and metabolism in vivo has been a long timewish of both researchers and clinicians alike. The recentdevelopment of the first technique for measurement ofmitochondrial PO2 (mitoPO2) in intact cells (1) has beendemonstrated to be a big step forward to potentialachievement of this goal. This optical technique is basedon measurement of the oxygen-dependent lifetime of thedelayed fluorescence of aminolevulinic acid (ALA)induced protoporphyrin IX (PpIX). The method hasproven to be useful in vivo and has now been validatedboth in liver (2) and heart (3).

Methods: Principle of mitoPO2 measurement byoxygen-dependent quenching of ALA enhanced PpIX.(a) Principle by which ALA administration enhancesmitochondrial PpIX levels. ALA, 5-aminolevulinic acid;PBG, porphobilinogen; UPIII, uroporphyrinogen III;CPIII, coporporphyrinogen III; and PpIX, protopor-phyrin IX. (b) Jablonski diagram of states and state tran-sitions of PpIX and its interaction with oxygen. S0, S1,and S2 represent the ground state and first and secondexcited singlet states, respectively. T0, T1, and T2 repre-sent the ground (triplet) state and first and second excitedtriplet states, respectively. kq is the rate constant of T1

quenching by oxygen. (c) PpIX emits delayed fluores-cence after excitation by a pulse of green (510 nm) light.The delayed fluorescence lifetime is oxygen-dependentaccording to the Stern–Volmer equation (inset), in whichkq is the quenching constant and s0 is the lifetime at zerooxygen. (Adapted from reference 2)

Results/discussion: This presentation will look back at thedevelopmentof the technique, present somecurrentwork andwill discuss potential clinical application of the technology.

Declaration of interest: E.G. Mik is founder and share-holder of Photonics Healthcare B.V., a company thatholds exclusive licences to patents related to the discussedtechnology from both the Academic Medical CenterAmsterdam and the Erasmus Medical Center Rotterdam.

REFERENCES

1. Mik EG, Stap J et al. Mitochondrial PO2 measured bydelayed fluorescence of endogenous protoporphyrinIX. Nat Methods. 2006 Nov; 3(11): 939–45.

2. Mik EG, Johannes T et al. In vivo mitochondrial oxygentension measured by a delayed fluorescence lifetimetechnique. Biophys J. 2008 Oct; 95(8): 3977–90.

3. Mik EG, Ince C et al. Mitochondrial oxygen tensionwithin the heart. J Mol Cell Cardiol. 2009 Jun; 46(6):943–51.

26 Journal of Clinical Monitoring and Computing

21. ABSOLUTE EIT COUPLED TO A BLOOD GASPHYSIOLOGICAL MODEL FOR THE ASSESSMENT OF LUNGVENTILATION IN CRITICAL CARE PATIENTS

S. Mohamad-Samuri1, M. Mahfouf1, M. Denaı1,

J. J. Ross2 and G. H. Mills2

1Dept of Automatic Control and Systems Engineering, university of

Sheffield, Sheffield, UK; 2Dept of Critical Care and Anaesthesia,

Northern General Hospital, Sheffield, UK

Introduction: The authors propose to use a previouslydeveloped data-driven physiological model (SOPAVent[1]) for continuous and non-invasive blood gas predictionsin combination with the Sheffield Mk3.5 Absolute Elec-trical Impedance Tomography (aEIT) [2] system to assesslung functions and guide ventilation therapy in criticalcare patients (Figure 1).

Methods: In aEIT, the Mean End Expiratory lungVolume (MEEV) should have the ability to provideregional information on the patient’s lung behaviour. Tomodel the relationship between MEEV and the relevantventilator parameters, a series of clinical trials have beenconducted on five (5) ITU patients at the NorthernGeneral Hospital, Sheffield, UK. Two modelling tech-niques (neural networks (NN) and neural-fuzzy) havebeen applied in order to elicit such relationships which areof a nonlinear nature.

Results: Figure 2 shows the results of one clinical trialperformed on four successive days on the same ITUpatient. A decrease in the Peak End-Expiratory Pressure(PEEP) levels leads to decreased lung resistivity andMEEV which agrees with [3].

Finally, the clinical exploitation of the models is eval-uated by comparing the predicted blood gas information(PaO2

and PaCO2) obtained from SOPAVent and the

regional lung volume information (MEEV) provided bythe ANFIS model subject to changes in PEEP settings.Table 1 summarises these results.

Discussion: Mean end-expiratory lung volume (MEEV)calculated from aEIT is a feature parameter that revealsvolume of air present in the lungs at the end of patients’expiration. In this study, increasing PEEP has lead toincrease in MEEV (predicted from ANFIS model) andPaO2

(predicted from SOPAVent model). This correlationshows that both models are capable of providing infor-mation on patients’ lung behaviour in response to venti-lation therapy. These sets of information should lead to abetter understanding of phenomena surrounding venti-lated patients in order to support decision-making andguide ventilator therapy. However, more ventilated pa-tients EIT data are needed to further improve the accu-racy of MEEV prediction. Knowledge from experts will

also be included in the form of decision rules for sug-gesting adequate ventilator parameters settings.

REFERENCES

[1] A. Wang, M. Mahfouf and G.H. Mills, ‘‘A continu-ously updated hybrid blood gas model for ventilatedpatients,’’ The 6th IFAC Symposium on Modelling andControl in Biomedical Systems, Reims, France, 2006.

Fig. 1 Advisory system for the management of ventilated critical carepatients.

Fig. 2 Lung abolute resistivity and air volume measured by aEIT atdifferent PEEP levels on an ITU patient.

Table 1 MEEV, PaO2and PaCO2

predicted by the models fol-lowing PEEP changes

PEEP

(mmHg)

MEEV

(l)

PaO2

(mmHg)

PaCO2

(mmHg)

12.0 4.94 11.56 6.14

11.0 4.87 11.22 5.89

10.0 4.80 10.87 5.67

9.0 4.73 10.53 5.47

8.0 4.67 10.19 5.28

Abstracts 27

[2] A. J. Wilson, P Milnes, A R Waterworth, R HSmallwood and B H Brown, ‘‘Mk3.5: a modular,multifrequency successor to the Mk3a EIS/EITsystem,’’ Physiol. Meas., 22, 2001, 49–54.

[3] J. Hinz, O. Moerer, P. Neumann, T. Dudykevych,G. Hellige, M. Quintel (2005). ‘‘Effect of positiveend-expiratory-pressure on regional ventilation inpatients with acute lung injury evaluated by electricalimpedance tomography.’’ European Journal of Anes-thesiology, 2005, 22, 817–825.

22. DECISION-SUPPORT FOR CLINICIANS—HOW TOIMPLEMENT

Alan H. Morris, MD

Intermountain Healthcare and University of Utah, Salt Lake City,

United States

Introduction: I focus on clinical trial application of cliniciandecision-support because this is a first step in providingthe credible information necessary to build a foundationfor wide spread use of clinician decision-support in clinicalpractice. Meeting the scientific requirements of rigorousclinical trials (clinical experiments) highlights similarchallenges that exist in the usual clinical care practiceenvironment.

Compliance of physicians with evidence-based treat-ments or guidelines is low across a broad range of health caretopics, in part because we lack widespread application ofdetailed clinical decision-support protocols. This lowclinician compliance contributes to uneven cointerventioneffects in clinical trials and thus contributes to unnecessaryvariation of clinical trial results. Cointerventions are con-founders introduced after allocation of subjects to theclinical trial experimental groups. Cointerventions, unlikeconfounders present before randomization, cannot be madeuniform across clinical trial groups through randomization.Many cointerventions are clinical care processes thatinfluence clinical trial outcomes, independent of theexperimental clinical trial intervention under study.

Experimental method and result reproducibility isrequired before new information is included in standardsources in many scientific domains. This is a scale anddomain-independent scientific requirement. The absenceof detailed clinical decision-support protocols is a criticalbarrier to the uniform management of cointerventionsneeded to conduct high quality clinical trials (1, 2). Theclinical research community does not possess tools to

standardize clinician decisions associated with delivery ofcointerventions and cointerventions are not commonlycontrolled in clinical trials. As a result clinical trials, andespecially non-blinded clinical trials like those ofmechanical ventilation, suffer from excess variation, non-reproducible methods, low scientific credibility, and var-iable results (2, 3). Cointervention effects likely explainmany inconsistencies observed in different studies of thesame putative intervention. Much of the often inconsis-tent and conflicting results of clinical trials (4, 5) andclinical care are likely due to non-reproducible methodsbecause the judgments of clinicians become an unarticu-lated and unidentifiable part of the experimental or clin-ical care method. These unidentified and unarticulatedelements influence outcomes in different studies andclinical reports and remain a barrier to understanding.

Methods: We embed rules (intelligence) into theeProtocols to minimize avoidable errors and omitteddocumentation, and to maximize the use of best practices.As data are input into the system they trigger one or morerule sets; such rules may also be invoked by passage oftime. Output from the eProtocol decision logic is storedin the patient’s eProtocol database, and sent to theappropriate caregiver(s) at the bedside. We develop, val-idate, and establish safety of the eProtocols using maturemethods (1, 2, 6–8).

Results: We have built, validated, employed clinically,and distributed adequately explicit bedside computerprotocols (eProtocols) that enable reproducible clinicalcare in critical care medicine for mecfhanical ventilation,intravenous fluid, and blood glucose management(1, 2, 6–10). eProtocols are adequately explicit computerprotocols that enable reproducible clinician decisionmethods that can control experimental cointerventions.An adequately explicit protocol can elicit the same deci-sion from different clinicians when faced with the sameclinical information. Clinician compliance with oureProtocol recommendations is 94%.

Discussion: Adequately explicit computer protocolsenable a reproducible clinician decision method that stan-dardizes clinician decision making while retaining patient-specific treatment and preserving ultimate cliniciandecision-making authority (1, 2, 6, 8, 11). Individualizedpatient care is preserved because the computer protocolrequires explicitly, patient-specific, clinical data. Differ-ences in clinical data represent unique patient expressions ofthe disease. This leads to different and individualized rec-ommendations from the computer protocol for eachpatient, even though the decision-making logic is the samefor all patients. Therefore, eProtocols enable a reproducible

28 Journal of Clinical Monitoring and Computing

clinician decision method that is adaptive, responds topatient changes, and individualizes patient care decisions.

REFERENCES

1. Morris A. Developing and implementing computer-ized protocols for standardization of clinical decisions.Ann Intern Med. 2000;132: 373–83.

2. Morris A. The importance of protocol-directed patientmanagement for research on lung-protective ventila-tion. In: Dreyfuss D, Saumon G, Hubamyr R, editors.Ventilator-induced lung injury. New York: Taylor &Francis Group; 2006. p. 537–610.

3. Wennberg JE. Unwarranted variations in healthcaredelivery: implications for academic medical centres.BMJ. 2002 October 26, 2002;325(7370): 961–4.

4. Singh JA, Hodges JS, Toscano JP, Asch SM. Quality ofcare for gout in the US needs improvement. ArthritisRheum. 2007 Jun 15;57(5): 822–9.

5. Wiener RS, Wiener DC, Larson RJ. Benefits and risksof tight glucose control in critically ill adults: a meta-analysis. JAMA. 2008 Aug 27;300(8): 933–44.

6. Morris A, Orme Jr J, Truwit J, Steingrub J, Grissom C,Lee K, et al. A replicable method for blood glucosecontrol in critically ill patients. Crit Care Med. 2008Jun;36: 1787–95. PMID: 18520641.

7. Morris AH, Orme J, Rocha BH, Holmen J, Clemmer T,Nelson N, et al. An Electronic Protocol for Translation ofResearch Results to Clinical Practice: A PreliminaryReport. J Diabetes Sci Technol. 2008 September2008;2(5): 802–8.

8. Thompson B, Orme J, Zheng H, Luckett P, Truwit J,Willson D, et al. Multicenter Validation of aComputer-based Clinical Decision Support Tool forGlucose Control in Adult and Pediatric Intensive CareUnits. J Diabetes Sci Technol. 2008 May 2008;2(3):357–68.

9. East T, Heermann L, Bradshaw R, Lugo A, Sailors R,Ershler L, et al. Efficacy of computerized decisionsupport for mechanical ventilation: Results of a pro-spective multi-center randomized trial. Proc AMIASymp. 1999: 251–5.

10. Sorenson D, Grissom CK, Carpenter L, Austin A,Sward K, Napoli L, et al. A frame-based representa-tion for a bedside ventilator weaning protocol.J Biomed Informatics. 2008 Jun;41(3): 461–8.

11. Sward K, Orme J, Jr., Sorenson D, Baumann L,Morris AH. Reasons for declining computerizedinsulin protocol recommendations: application of aframework. J Biomed Inform. 2008 Jun;41(3):488–97.

23. A LINUX-BASED ANAESTHESIA WORKSTATION

R. W. D. Nickallsa, S. Dalesb, A. K. Nicec

aDepartment of Anaesthesia, City Hospital Campus, Nottingham

University Hospitals, Nottingham, UK(Email: dick@nickalls.org);bPurrsoft, Oxford, UK; cDepartment of Information and Com-

puting Technology, City Hospital Campus

Introduction: This talk describes our experience of devel-oping a Linux-based anaesthesia information managementsystem (AIMS) as part of a research program [1]. During2003–2004 our original MS-DOS program [2] wasrewritten for the Linux operating system using open-source tools. Electrical safety was overseen by theDepartment of Medical Physics. The default screen gives acontinuous trend display of measured and derivedparameters. Pull-down menus allow the inputting of drugsand events via mouse and/or keyboard entries. The sys-tem also offers ‘help’ and some decision support, andautomatically prints out the Anaesthesia Record at the endof the operation in a form suitable to be placed directlyinto the clinical notes. An example screenshot and MAC-widget detail is shown in Figure 1.

Materials and methods: Data acquisition and displaymodule: This is written in C/C ++ and uses the QtGUI library (standard with Linux systems). Serial-portdata was accessed at 5 s intervals from a Datex S/5anaesthesia monitor and displayed in both trend andtabular formats on the screen, and saved to disk. It uses amodular driver system so that new monitor types can beadded. Its configuration script system permits screenlayouts and other parameters to be easily customised tolocal requirements.

MAC display widget: In order to better gauge depth ofanaesthesia [3, 4] the screen display incorporates a real-time age-corrected MAC display widget (Figure 1),which is positioned in the lower right part of the maindisplay screen. The MAC-widget displays the currentMAC value, and implements an alerting colour change(to red) to warn of an out-of-range value, and hencegreatly facilitates the avoidance of inadvertent awarenessof the patient under anaesthesia.

Diabetes alert module: This is a Perl script which makesuse of the Linux Kalarm utility. Tk widgetsare used topresent a menu which allows the user to quickly setspecial alerts to prompt regular monitoring of bloodglucose. A ‘help’ system allows the user to access protocolsfor the insulin management of diabetic patients duringmajor surgery.

Drug-menu module: This is a pull-down drug menusystem which incorporates the standard NHS Dictio-nary of Medicines and Devices (DM + D) EU drug-list

Abstracts 29

database [5]. The NHS listing currently consists of about2000 drug names and preparations, and is updated weekly.Of the various subsets of the NHS drug database whichare available for download we found the Virtual Thera-peutic Moieties (VTM) files to be the most useful, sincethese held the comprehensive list of standardised drugnames and little else. Since this is a very large down-loadable XML database, we wrote a Perl script to auto-matically convert the XML files into a format suitable forus to easily import into our system, with modifications asnecessary.

Printing and data output module: Data output is gen-erated using Perl scripts which coordinate data manipula-tion, graph plotting (using GNUplot), and typesetting(using LaTeX) [6]. Two output formats are used, as follows:

1. A paper Anaesthetic Record suitable for the patientnotes, including the graphic trends and keyboard entries(events, procedures, drugs given, etc.), which is printedin the operating theatre at the end of anaesthesia.

2. A comprehensive HTML format allowing easy access tothe data files, graphs, programs and coordinating scripts.

Results/discussion: This Linux prototype started beingused clinically in the thoracic operating theatre in 2004,and was found easy to use by both consultant and traineeanaesthetists. It was extremely useful both clinically andmedico-legally.

REFERENCES

1. Nickalls, RWD and Ramasubramanian R. (1995).Interfacing the IBM-PC to medical equipment: the art

of serial communication. ISBN 0-521-46280-0;pp 402 (Cambridge University Press).

2. Nickalls RWD (1996). An automated AnaesthesiaRecord System using free text-based software. 16thInternational Symposium on Monitoring and Com-puting in Anaesthesia and Intensive Care (Rotterdam,Holland; May 1996).

3. Nickalls RWD and Mapleson WW (2003). Age-related iso-MAC charts for isoflurane, sevoflurane anddesflurane in man. British Journal of Anaesthesia;91 (August), 170–174.

4. Nickalls RWD and Mahajan R (2010). Awareness andanaesthesia: think dose, think data. British Journal ofAnaesthesia; 104, 1–2.

5. www.dmd.nhs.uk/dictionary/6. Nickalls RWD (1998). TeX in the operating theatre:

an Anaesthesia application. TUGboat; 19 (3), 7–9.http: //www.tug.org/TUGboat/Articles/tb19-3/tb60nick.pdf

24. A HOSPITAL FOR POST-ICU PATIENTS ON LONG TERMMECHANICAL VENTILATION IN JAPAN

A. Okamura, T. Ishitani, M. Fukuda,

T. Yamamura

Heiseikai InoUE hospital, S7W2 Chuo-ku, Sapporo 064-0807

Japan

Introduction: Emergency medicine and intensive care havefacilitated the increased survival of severely ill patients.However, among these survivors, some are difficultto wean from ventilators. In the United States, these

Fig. 1 Trend-data screenshot (left) and real-time age-corrected MAC-widget display (right) of data from a Datex S/5 monitor. If the age-corrected MAC fallsoutside the set valid range it triggers both audible and visual alarms (turns the dial of the MAC-widget red). (images�Nickalls and Dales 2001–2010).

30 Journal of Clinical Monitoring and Computing

ventilator-dependent patients are treated at Long TermAcute Care (LTAC) Hospitals1)2). In Japan, such hospi-tals have not yet been organized. Therefore, we createdthe first of this kind of hospital in Japan.

The estimated total number of the long-term venti-lator-dependent patients in Japan is approximately15,800. Among these, 2,000 have neuro-musculardiseases and are mainly treated in national hospitals.Another 10,000 are on home mechanical ventilation.The remaining 3,800 post-ICU ventilator-dependentpatients are treated in ordinary hospital wards. Webelieve that a hospital that specializes in post-ICUmechanical ventilation is necessary to improve the out-come of the ventilator dependent patients and to mini-mize possible incidents/accidents.

Methods: We acquired a private hospital with 82 beds in2003, and renovated it to our purpose.

The hard-ware added during the renovation included:ventilators, patient monitors, central monitors, oxygenand artificial air gas supplies, co-generation power suppliesfor ventilators and monitors, and a patient informationsystem (receipt claiming, medical records, radiologyimages, and laboratory data).

We have organized special teams, including thosefocused on risk management, infection control, medicalrecords, education, pressure ulcer treatment, and processKAIZEN (improvement). Physicians, nurses and otherhospital staff members worked together with these teamsto improve the processes throughout the hospital.

Results: Within the first year of opening, we took careof 35 ventilator-dependent patients. Since 2007, we haveconstantly been taking care of 75 ventilator-dependentpatients. In the meantime, the patient-bed occupancy ratehas risen to almost 100%. The successful weaning rate was31% which was similar to the result reported by Carsonet al3) (35%) but lower than the results of a multicenteroutcome study4) (54%). The ADL (activity of daily life) ofthe patients was evaluated by a score from1 (bed-bound)to 5(able to walk). The average ADL score slightlyimproved in our patients from 1.6 (on admission) to 2.1(on discharge/in March 2009). With regard to the staffeducation, increased compliance with standard precau-tions decreased the incidence of hospital-acquired infec-tions, and resulted in spending cuts of antibiotics (from\300 million in 2004 to \100 million in 2008). Ourmicrobiological study of tracheal suction catheters per-mitted the reuse of the suction catheter within 12 h underthe circumstances of our strict suction manual. We havefound many other divergent processes and have workedwith the teams to improve the efficiency of the processesin the hospital. These staff education methods and clinicalstudies at our hospital have decreased wasteful expendi-tures and increased the process efficiency, finally resulting

in increased ordinary profit. We had 500 to 600 reports ofincidents and accidents each year. One of the indices ofthe risk management activity is the (number of reports)/(number of beds) ratio. Our LTAC ratio of 6.9 is higherthan that of non-LTAC hospitals 4.4 in Japan. However,during its 6 years in operation, there have been no acci-dental ventilator-related patient deaths in our hospital. Theordinary profit rate is currently 16% which is higher thanthe average rate of Japanese private hospitals (6%). Wewere certified by the Japan Council for Quality HealthCare (JCQHC) Version 5 which assures the functionalquality of the hospital in 2007. The national tax adminis-tration accredited our hospital with tax reduction for thetransparent business management in 2009.

Discussion: The problem with prolonged mechanicalventilation (PMV) in Japan is that there is no specializedhospital for post-ICU PMV. There is no data about post-ICU PMV in Japan, so we cannot compare the successfulweaning rate out-side the hospital in Japan. The differencein the successful weaning rate between our hospital andthe rate reported by Scheinhorn comes from the patientdemographics and exclusion criteria (end-of-life care,terminal weaning and so on). If we adopted the sameexclusion criteria, our successful weaning rate would be52%. The improvement of ADL also seemed to be slight,because of the involvement of difficult-to-wean patients.Our efforts to improve the successful weaning rateinclude: respiratory physical therapy, CASS (continu-ous aspiration of subglottic secretions) for preventingVAP, and nutrition management using a metabolic ana-lyzer during mechanical ventilation to measure basicenergy expenditure (excessive caloric intake results inhypercapnia).

Our enterprise proved that the hospital is useful notonly for the care of post-ICU ventilator-dependentpatients but also is economically feasible. Continuouseducation of hospital staff and process KAIZENseem to be the key elements for the success of thisenterprise.

REFERENCES

1. Seneff MG et al. The impact of long-term acute-carefacilities on the outcome and cost of care for patientsundergoing prolonged mechanical ventilation. CritCare Med. 28(2): 342–350, 2000

2. Alexander CW, et al. Prolonged mechanical ventila-tion. Review of care settings and an update on pro-fessional reimbursement. Chest. 133(2): 539–545,2008

3. Carson SS, et al. Outcomes after long-term acute care.An analysis of 131 mechanically ventilated patients.

Abstracts 31

Am J Respir Crit Care Med. 159(5 Pt 1): 1568–1573,1999

4. Scheinhorn DJ et al. Post-ICU mechanical ventilationat 23 long-term care hospitals. A multicenter outcomestudy. Chest 131 (1): 85–93, 2007

25. THE NEXFIN – A NEW NON-INVASIVE MONITOR FOR THEMEASUREMENT OF CONTINUOUS CARDIAC OUTPUT

Azriel Perel, MD, Professor and Chairman

Department of Anesthesiology and Intensive Care, Sheba Medical

Center, Tel Aviv University, Tel Aviv, 52621 Israel

The Nexfin HD monitor (BMeye, Amsterdam; http: //www.bmeye.com) measures cardiac output (CO) contin-uously in a totally non-invasive manner by an inflatablefinger cuff which is the only interface with the patient. TheNexfin HD measures continuous finger blood pressure(BP) by the Volume Clamp Technology and transforms itinto a brachial artery waveform. Applying a 3rd third gen-eration pulse contour method which is based on a full 3-element Windkessel model, and following the input ofpatient’s gender, age, height and weight, continuous CO(CCO) is measured and displayed. The CCO is calculatedwithout external calibration although it can be calibratedexternally. This technology is based on an extension andcombination of elements of two previous generations ofalgorithms, the so-called corrected characteristic imped-ance or cZ method (Wesseling 1974) and the Modelflowmethod (Wesseling 1993). The parameters that are mea-sured by the Nexfin HD include continuous BP (systolic,diastolic, mean), heart rate, continuous cardiac output(CCO), stroke volume (SV), systemic vascular resistance(SVR), and left ventricular contractility (dP/dT).

The truly non-invasive nature of the Nexfin HDallows the measurement of CCO in a much widervariety of patients than was hitherto possible. Originally,the Nexfin HD was introduced in Cardiology clinics forthe performance of tilt-test for the detection of ortho-static hypotension. A recent study using the Modelflowhas shown that the early postoperative postural cardio-vascular response is impaired after radical prostatectomywith a risk of orthostatic intolerance, limiting earlypostoperative mobilization. Both the tilt test and the sit-stand test take advantage of the fact that the Nexfin HDprovides real-time CCO, allowing the detection of theimmediate response to diagnostic and therapeutic chal-lenges. These include passive leg raising, fluid challenge,start of inotropes, optimization of cardiac resynchroni-zation therapy, etc. Indeed, the continuous real-time

CO measurement by this uncalibrated method may bemore useful and provide more accurate informationabout the changes in CO than intermittent CO mea-surements by thermodilution with their inherent vari-ance.

One of the most interesting areas where the potential ofthe Nexfin HD can be fully expressed is perioperativecare. It is well recognized that a small group of patientsaccount for the majority of peri-operative morbidity andmortality. These ‘high-risk’ patients have a poor outcomedue to their inability to meet the oxygen transport de-mands imposed on them by the nature of the surgicalresponse during the peri-operative period. It has beenshown that by targeting specific hemodynamic and oxy-gen transport goals at any point during the peri-operativeperiod, the outcomes of these patients can be improved.Most studies on perioperative optimization have usedrepetitive fluid challenges in order to maximize the CO.CO was measured in most of these studies by theesophageal Doppler and by the FlowTrac. However, theesophageal Doppler can be used only after induction ofanesthesia, while the FlowTrac necessitates the presenceof an arterial line. The non-invasive nature of the NexfinHD and its semi-disposable finger probe make the use ofthis monitor to seem ideal in this important setting, andthe preliminary results are very promising. In fact, theNexfin is perfectly suitable for any patient who is sick or atrisk enough, in terms of hemodynamic instability, towarrant the need for continuous real-time hemodynamicmonitoring, but who is not sick enough to warrant theuse of invasive lines and catheters with their associatedcomplications.

26. GLUCOSAFE—A MODEL-BASED MEDICAL DECISIONSUPPORT SYSTEM FOR TIGHT GLYCEMIC CONTROLIN CRITICAL CARE

Ulrike Pielmeiera

aCenter for Model-based Medical Decision Support, Aalborg Uni-

versity, Fredrik-Bajers-Vej 7, 9220 Aalborg, Denmark

Introduction: Hyperglycemia during critical illness is com-mon and is associated with increased mortality, morbidityand prolonged stay in intensive care [1][2]. The past decadehas seen many attempts to improve survival by regulatingblood glucose using intensive insulin therapy (IIT) pro-tocols [3][4]. However, consistent control has provenelusive, not least because typical IIT protocols ignore thecarbohydrate intake of patients [5]. An effective methodthat achieves and maintains ‘‘tight’’ blood glucose lev-els (i.e. in the range from 4 to 6 mmol/l) without high

32 Journal of Clinical Monitoring and Computing

glucose variances and without increasing insulin-inducedsevere hypoglycemia (< 2.2 mmol/l) has yet to emerge [6].

This work assesses the effectiveness of the computerizeddecision support system ‘‘Glucosafe’’ for tight glycemiccontrol in critical care. This system advises insulin therapyand infusion rates of enteral and parenteral nutrition, basedon blood glucose predictions with a physiological insulin-glucose model and patient-specific data [7]. Pilot testingshows significant improvements of glycemic control in aprospectively controlled cohort of intensive care patients[8]. In a retrospective analysis of the pilot study data themodel is assessed with regard to how accurately bloodglucose was predicted, and whether the predictive accu-racy can be improved by two physiological model exten-sions, regarding the decreased delivery rate of nutrientsthat is often observed in critical care patients with delayedgastric emptying [9], and the dependency of pancreaticinsulin secretion on the blood glucose level [10].

Methods: The blood glucose concentrations of 10hyperglycemic patients admitted to a neuro- and traumaintensive care unit were retrospectively predicted using a)the original Glucosafe model [7] b) the Glucosafe modelincluding a feedback loop between blood glucose andpancreatic insulin secretion rate c) the Glucosafe modeland a reduced rate of appearance of enterally administerednutrition in the intestinal reservoir d) both extensions asdescribed in b) and c). Prediction errors were expressed asabsolute percent error (APE) from measured concentra-tions; the comparison was based on median APEs fordifferent prediction time lengths, reflecting intervalsbetween measurements of up to 5 h.

Results: The model predictive accuracy improvedmodestly for each one of the two model extensions. Thegreatest reduction in prediction error was achieved whenboth model extensions were included in the Glucosafemodel. For predictions time lengths (in hours) of 0.5–1.5 h, 1.5–2.5 h, 2.5–3.5 h, 3.5–4.5 h and 4.5–5.5 h, themedian APE was 9.7%, 11.2%, 14.8%, 15.1% and 17.7%with the Glucosafe model, compared to 9.2%, 10.1%,12.3%, 13.2% and 16.6% with both of the model exten-sions included, for the same prediction time lengths.

Discussion: Predicted blood glucose concentrations withthe Glucosafe model in its original form [7] are sufficientlyaccurate for typical time intervals between two measure-ments. The pilot trial results [8] showed that glycemic controlwas significantly improved, while no hypoglycemic eventwas observed. Thus, model-based predictive control basedon the Glucosafe model may be a step towards a consistentreduction of elevated blood glucose levels. This retrospectiveanalysis also explored two physiological model extensions,which modestly improved the model’s predictive accuracy.However, as the data used in this study were from a smallcohort of patients with similar admission diagnosis, groups of

other patients with a different disease background should beused to verify these preliminary results.

REFERENCES

1. Falciglia M, Freyberg RW, Almenoff PL, et al.Hyperglycemia-related mortality in critically illpatients varies with admission diagnosis. Crit CareMed 37 (12): 3001–3009, 2009.

2. Krinsley JS: Association between hyperglycemia andincreased hospital mortality in a heterogeneous popu-lation of critically ill patients. Mayo Clinic Proceedings78 (12): 1471–1478, 2003.

3. Meijering S, Corstjens AM, Tulleken JE, et al.Towardsa feasible algorithm for tight glycaemic control incritically ill patients: a systematic review of the litera-ture. Critical Care 10 (1): R19, 2006.

4. Chase JG, Hann CE, Shaw GM, et al. An overview ofglycemic control in critical care - relating performanceand clinical results. Journal of Diabetes Science andTechnology 1 (1): 82–91, 2007.

5. Kalfon P, Preiser JC. Tight glucose control: should wemove from intensive insulin therapy alone to modu-lation of insulin and nutritional inputs? Crit Care 12:156, 2008.

6. Dossett LA, Collier B, Donahue R, et al. Intensiveinsulin therapy in practice: Can we do it? J ParenterEnteral Nutr 33 (1): 14–20, 2009.

7. Pielmeier U, Andreassen S, Nielsen BS, et al. A simu-lation model of insulin saturation and glucose balance forglycemic control in ICU patients. Computer Methodsand Programs in Biomedicine 97 (3): 211–222, 2010.

8. Pielmeier U, Andreassen S, Juliussen B, et al. TheGlucosafe system for tight glycemic control in critical care:A pilot evaluation study. J Crit Care 25 (1): 97–104, 2010.

9. Chapman M, Fraser R, Matthews G, et al. Glucoseabsorption and gastric emptying in critical illness. CritCare 13 (4): R140, 2009.

10. Polonsky KS, Given BD, Van Cauter E. Twenty-four-hour profiles and pulsatile patterns of insulinsecretion in normal and obese subjects. J Clin Invest81: 442–448, 1988.

27. IDENTIFYING ERGONOMIC REQUIREMENTS OF ICT FORHEALTHCARE WORKING SYSTEMS

Podtschaske, B., Aciksoz-Tavasli, F., Friesdorf, W.

Department for Human Factors Engineering & Product Ergonomics,

Technical University of Berlin, Germany

Abstracts 33

Introduction: In an ageing society solutions are needed,which can help to overcome the trade-off between theincreasing demands and the limited resources. An eligibleapproach to gain higher quality, safety, and efficiency inhealthcare working systems seems to be the use of moderntechnology such as information and communicationtechnology (ICT) (Carayon & Friesdorf, 2006).

Available ICT solutions for healthcare working sys-tems often lack in usability or evident process support.Errors can occur leading to inefficiency or even jeop-ardize the patient’s safety (Backhaus, 2010, Backhaus &Friesdorf, 2007). Different studies show that the antici-pated benefit (improving quality of care and reducingcosts) could be realized only partially (Himmelsteinet al., 2010). In most cases ICT is either partly or not atall compatible to other digital systems. Overall theprevalence rate of available (medical) information systemsas well as their range of functions is still in less satis-faction (Haas, 2005).

Due to medical and technical progress the way howICT can be used to support treatment processes arevarying strongly and changing quickly. For example, theimportance of applications used at home is continuouslyincreasing. Therefore the group of potential users (e.g.patient or care giving relatives) is expanding and thecustomer’s requirements are getting more heterogeneous.These aspects cause that the complexity in a user-orienteddevelopment process is more and more increasing.Therefore the ergonomic design of systems, products andservices matching the user demands is a highly challengingprocess especially for healthcare working systems. (Glende& Podtschaske & Friesdorf, 2009).

Methods: General guidelines have to be specified byusing usability context analyses. Guidelines on how toperform a usability context analysis are available in thestandards but they are insufficient for analyzing complexworking systems. In this case domain specific modelsand methods are necessary. The Objective of this con-tribution is to illustrate the issues related to the ergo-nomic development of ICT applications for healthcareworking systems. In order to solve these problemsdomain specific models and developed methods areshown to specify usability context analyses for healthcareworking systems. The presentation is divided into thefollowing parts:

1. Introduction of the general ergonomics approach andcorresponding International Standards;

2. Demonstration of characteristics of the field ‘‘health-care working systems’’;

3. Development of domain specific ergonomic modelsand methods for an usability context analysis inhealthcare working systems;

4. Evaluation of the developed models with the help of acase study.

Results: The case study illustrates various tasks thathave to be supported by ICT during an abdominalsurgical treatment process. With the aid of a require-ment catalogue it is possible to derive detailed functionsthat are suitable for the task. Such a catalogue ofrequirements is essential in order to define and developan utilizable ICT for healthcare working systems. Somerequirements cannot be implemented easily. Forexample, the consideration of different demands ofstakeholders is not a trivial task. Design requirements,e.g. concerning aggregation of data and information,depend on the corresponding care provider and the(sub-) task.

The case study shows the suitability of models toillustrate complex interrelations of patient treatment.Based on the results the requirements are defined andtask-supporting functions are derived. The models affordthe development of a ‘‘common ground of understand-ing’’ between the potential user and the developersof the product. Necessary expertise and different per-spectives can be submitted in the process of productdevelopment and product evaluation. This is a crucialpre-condition for developing integrated and utilizable(software) products and therewith for developing ergo-nomic working systems.

Discussion: In order to improve the quality of the resultsfurther usability context analyses (e.g. chronic heart dis-eases, dementia patients, diabetes mellitus) are recom-mended. Additionally investigations such as usability testsand simulations are necessary to evaluate the efficiency ofuse and the satisfaction with the product.

REFERENCES

1. Backhaus, C. (2010). Usability Engineering in derMedizintechnik. Grundlagen, Methoden, Beispiele.Berlin, Heidelberg: Springer Verlag.

2. Backhaus, C. & Friesdorf, W. (2007). Human Factorsand Ergonomics in Intensive Care. In Carayon, P.(Ed.), Handbook of Human Factors and Ergonomicsin Healthcare and Patient Safety (pp. 835–850).Mahwah, New Jersey, London: Lawrence ErlbaumAssociates, Publishers.

3. Carayon, P. & Friesdorf, W. (2006). Human Factorsand Ergonomics in Medicine. In Salvendy, G. (Ed.),Handbook of the Human Factors and Ergonomics (pp.1517–1537). New Jersey: John Wiley & Son.

34 Journal of Clinical Monitoring and Computing

4. Glende, S. & Podtschaske, B. & Friesdorf, W. (2009).Senior User Integration: Ein ganzheitliches Konzeptzur Kooperation von Herstellern und alteren Nutzernwahrend der Produktentwicklung. In VDE/VDI-IT &BMBF (Eds.), Ambient Assisted Living: Technologien,Anwendungen. 2. Deutscher AAL-Kongress. Berlin,Offenbach: VDE Verlag.

5. Haas, P. (2005). Medizinische Informationssystemeund elektronische Krankenakten. Berlin, Heidelberg,New York: Springer Verlag.

6. Himmelstein, D.U. & Wright, A. & Woolhandler, St.(2010). Hospital Computing and the Costs and Qualityof Care: A National Study. The American Journal ofMedicine, 123 (1), pp. 40–46

28. SAFETY THROUGH STANDARDISATION

Dr M. S. Read, MBBS FRCA

Consultant Anaesthetist/Intensivist, University Hospital of Wales,

Cardiff, CF14 4XW, UK

Ideally, healthcare should be by Integrated DeliverySystems, but in practice it often has the characteristics ofa Complex Adaptive System. Trainee doctors are givenan extensive knowledge base and then an apprenticeship,and are then expected to do sensible things in a varietyof complex situations. However, most of what we do,we have done before. For these situations, pre-plannedstructured care is preferable. With good support struc-tured care will evolve as the organisation learns from pastmistakes. The Institute of Healthcare Improvement hassuggested that the research agenda in medicine shouldinclude, ‘‘How to bring engineering science into healthcare to improve care processes.’’ [1]. Quality has beensaid to comprise: Leadership; Policy & Strategy; People;Resource Management; Processes; and Results [2]. Thistalk focuses on people and processes, and their interac-tion: in other words, teamwork. Teaching this is a newscience, known in aviation as ‘‘Team Training’’, inmanagement as ‘‘Relational Co-ordination’’, and inmedicine as ‘‘Non-technical skills training’’. (‘‘ANTS’’isAnaesthesia Non-technical Skills Training.) Principlesinclude; mutual respect; shared knowledge; commongoals; constructive criticism; and role-based (not indi-vidual) working [3]. In Amalberti’s paper, ‘‘Five SystemBarriers to Achieving Ultrasafe Health Care’’[4], steps 2and 3 are abandonment of professional autonomy andtransition from the mindset of craftsman to that of anequivalent actor. Teams share processes, which musttherefore be formalised. Examples of structured care

methods (SCMs) include integrated care pathways,guidelines, protocols, algorithms, care bundles, andtreatment order-sets. Once SCMs are commonly used,unexplained variation in practice can be recognised to bea bad thing of itself and should be eliminated. MostSCMs are benefits: the complexity of healthcare will beeasier to manage [5]; there should be less human effortthan using paper systems; it is a short intellectual step tousing them to command healthcare rather than simply todescribe it; and equipoised ideas can be compared byrandomisation to competing pathways. The next gen-eration of doctors will be directing this activity, so thepresent generation of trainees should be being trainedin it.

REFERENCES

1. Grol R, Berwick D, Wensing M. On the Trail ofQuality and Safety in Health Care. BMJ. 2008 Jan12;336(7635): 74–6.

2. http: //www.efqm.org/en/3. Gittell, Jody Hoffer PhD et al. Impact of Relational

Coordination on Quality of Care, Postoperative Painand Functioning, and Length of Stay: A Nine-HospitalStudy of Surgical Patients. Medical Care. 38(8): 807–819, August 2000.

4. Rene Amalberti et al. Improving Safety By ChangingProcesses. Five System Barriers to Achieving UltrasafeHealth Care. Ann Intern Med. 2005;142: 756–764.

5. Dr. David Bates & Dr. Atul Gawande. ImprovingSafety with Information Technology. New EnglandJournal of Medicine. 348: 2526–34. 2003

29. THE CURRENT STATUS OF THE AUTOMATIC LUNGPARAMETER ESTIMATOR

Lars P. Thomsen1, Bram W. Smith1, Søren

Kjærgaard2, Per Thorgaard2, Egon Toft3, Steen

Andreassen1, and Stephen E. Rees1

1Center for Model Based Medical Decision Support System and3Department of Health Science and Technology, Aalborg Univer-

sity, Fredrik Bajers Vej 7, DK-9220 Aalborg, Denmark;2Departments of Anesthesiology, Aalborg Sygehus, Aarhus Uni-

versity Hospital, Aalborg, Denmark, DK-9000

Introduction: Patients with pulmonary gas exchangeabnormalities are at risk of developing hypoxaemia andhypercapnia. The underlying cause of the impaired gasexchange is due to mismatch between ventilation andperfusion of the lungs, ranging from V/Q equals zero i.e.

Abstracts 35

pulmonary shunt, to infinitely high V/Q i.e. alveolar deadspace.

In 2002 a system was described for measuring pul-monary gas exchange properties. The system known asthe Automatic Lung Parameter Estimator (ALPE) [1] isillustrated in Figure 1i, and consisted of a ventilator (A), agas analyser with pulse oximeter (B), and a computerdisplaying the captured and recorded data (C). ALPE isused to conduct a 10–15 min procedure where the frac-tion of inspired oxygen (FiO2) is varied in four to sixsteps, such that the peripheral oxygen saturation (SpO2) isin the range 88–100%. ALPE fits these data to a mathe-matical model to obtain parameters describing pulmonaryshunt and V/Q mismatch.

Methods: Clinical evaluation of ALPE has lead to animproved understanding of the systems limitations, the majorbeing that ALPE includes a mechanical ventilator. Potentialexisted for simpler, cheaper technology for performingALPE in patients not undergoing mechanical ventilation.This abstract describes these limitations and addresses thedevelopment of a commercial version of ALPE.

Results: The use of ALPE has lead to development ofdedicated versions for mechanically and spontaneouslybreathing patients respectively. The research version formechanically ventilated patients remains in a form similarto that of the 2002 version, having been updated forrecent communication protocols and the latest readilyavailable hardware. For spontaneously breathing patients acommercially available version of ALPE [2] has beendeveloped (Figure 1ii & iii). Gas delivery and measure-

ments are integrated into a single respiratory unit (Fig-ure 1iii). The patient breathes freely through the unitwhich is open to atmospheric air, introducing minimalresistance. Flow and oxygen fractions are measured viasensors integrated into the respiratory unit (a and b inFigure 1iii) with sensors protected by an antibacterial andhumidity filter (d in Figure 1iii). A small amount of eitheroxygen or nitrogen is injected into the inspiratory stream(c in Figure 1iii). This mixing achieves inspired oxygenfractions ranging from 15% to 40%. The oxygen satura-tion of blood is measured with an integrated pulseoximeter.

Discussion: The commercially developed version ofALPE ensures a fast and precise description of pulmonarygas exchange in spontaneous breathing patients withoutthe need for a mechanical ventilator or other additionalequipment.

REFERENCES

1. Rees SE, Kjaergaard S, Thorgaard P, Malczynski J,Toft E, Andreassen S. The automatic lung parameterestimator (ALPE) system: non-invasive estimation ofpulmonary gas exchange parameters in 10–15 min.J.Clin.Monit.Comput. 2002 Jan;17(1): 43–52.

2. Mermaid Care. Available at: http: //www.mermaidcare.dk/. Accessed 3/30/2010, 2010.

Fig. 1 Part i of the figure (from [1] with permission) shows the 2002 experimental version of ALPE. A ventilator (A1&A2), a gas analyser with oximeter(B1&B2), and a computer (C) collecting the data are seen. Also the gas-inlet (D&E) and the face mask (F) are shown. In part ii of the figure the commercialavailable ALPE essential is seen, consisting of the main unit including gas tanks and a respiration unit. Part iii shows a schematic drawing of the respirationunit where flow (a) and oxygen fraction (b) is measured by mainstream sensors and proper mixing of oxygen and nitrogen is ensured by injection occurringagainst the stream and being distributed via a lattice (c). The disposable tube also has an antibacterial and humidity filter (d).

36 Journal of Clinical Monitoring and Computing

30. COMPONENTS OF ANAESTHESIA INFORMATION SYSTEMS

N. van Schagen

Department of Anesthesiology, VU University Medical Center,

Amsterdam, The Netherlands

The department of anesthesiology of the VUmc hasdesigned and constructed its own Anesthesia InformationManagement System (AIMS) that is deployed on all 16OR’s.

The origin’s date back to 1992. The base of the AIMSis formed by a full disclosure flightrecorder. The system isstill being developed and extended.

A modern AIMS should in our opinion not onlymanage information but also be a platform to support newdevelopments and to implement new ideas. Without anydoubt it should be able to cope with the challenges of anever changing clinical and technical environment.

Some standard information techniques can help inbuilding such a system.

Of a number of strategies we would like to put emphasison principles that are most important in our view:

1) keep it simple (KIS)

2) build a modular system so one can use separatebuilding blocks.

One of the other challenges is interfacing with the userin a coherent way.

Web technology is a very handy tool for a consistentlook and feel while still being able to develop and alterwhat is ‘behind the scenes’. Open standards and opensource tools provide an excellent way of keeping thesystem flexible and up to date.

In the presentation we will show the modules involvedand we will also give a live demonstration of the featuresand possibilities of the complete system (Figure 1).

31. NEAR INFRARED SPECTROSCOPY (NIRS) TO MONITORTISSUE HAEMOGLOBIN (AND MYOGLOBIN) OXYGENATION

T. W. L. Scheeren

Department of Anaesthesiology, University Medical Center

Groningen (UMCG), University of Groningen, The Netherlands

Introduction: Tissue oxygenation may be monitored non-invasively by near infrared spectroscopy (NIRS) both onthe thenar eminescence (muscle) and on the forehead(brain). Thenar measurement have been used to guidetherapy in trauma patients (1) and to determine 5theprognosis of septic patients (2). The cerebral measure-ments have been shown beneficial in managing patients atrisk of cerebral ischemic injury (3), e.g. during cardio-pulmonary bypass (4). Up to now no information existson the value of NIRS monitoring during general surgery.Tissue oxygenation may be jeopardized during high-risksurgery, particularly in high-risk patients. We comparedintraoperative tissue oxygenation obtained in high andlow risk patients undergoing high and low risk surgery.

Methods: NIRS is based on the fact that oxygenated anddesoxygenated hemoglobin (and myoglobin) have differ-ent absorption spectra of near infrared light (700–950 nm). These can be determined at distinct wavelengthsby several commercially available NIRS devices such asInSpectra (muscle), Invos and Foresight (both cerebral). Inour study, tissue oxygenation was measured intraop-eratively on the thenar eminescence by near infrared

Fig. 1 Overview of the VUmc AIS system showing the black box (or flight recorder) as central element.

Abstracts 37

spectroscopy (StO2, InSpectra, Hutchinson Tech.) in atotal of 152 patients. Patients were stratified as high risk(HRP, ASA status ‡ III, age > 65 yr, n = 82) or low riskpatients (LRP, ASA status £ II, age < 65 yr, n = 70) andto receive high risk (HRS, tumor surgery, n = 121) orlow risk surgery (LRS, kidney transplantation n = 31).We defined a cut-off StO2 value of 80% to separatenormal tissue oxygenation (StO2 ‡ 80%) from tissue hy-poxia (StO2 < 80%).

Results: In cardiac surgical patients it has been shownthat a protocol aimed at increasing cerebral oxygen supplyrestored cerebral oxygenation assessed by NIRS in 84% ofcases (5). Similarly, the occurrence of StO2 values below50% increased the risk of cognitive dysfunction and pro-longed hospital stay threefold (6). In our study, tissuehypoxia occurred in 39% of HRP and in 19% of LRP atany time during surgery (p < 0.05, Wilcoxon test). Thedifference was even greater when looking at the occur-rence of more severe forms of tissue hypoxia (StO2 <70%): 8 vs. 3%, p < 0.01. In addition, tissue hypoxiaoccurred significantly more often in HRS compared toLRS (19 vs. 6%).

Discussion: Tissue hypoxia occurs frequently in theintraoperative setting, particularly in high-risk patientsundergoing high-risk surgery. In cardiac anaesthesia,where NIRS is more widely used, this technique mayoffer advantages in the prevention of cerebral hypoxia,e.g. during hypothermic circulatory arrest (7). Furtherstudies should look at the impact of intraoperative tissuehypoxia as well as of therapeutic interventions on post-operative patients’ outcome.

REFERENCES

1. Crookes BA, Cohn SM, Burton EA, et al. Noninva-sive muscle oxygenation to guide fluid resuscitationafter traumatic shock. Surgery 2004; 135 (6): 662–670.

2. Creteur J, Carollo T, Soldati G, et al. The prognosticvalue of muscle StO2 in septic patients. Intensive CareMed 2007; 33 (9): 1549–1556.

3. Murkin JM, Arango M. Near-infrared spectroscopy asan index of brain and tissue oxygenation. Br J Anaesth2009; 103 (Suppl 1): i3–13.

4. Vohra HA, Modi A, Ohri SK. Does use of intra-operative cerebral regional oxygen saturation moni-toring during cardiac surgery lead to improved clinicaloutcomes? Interact Cardiovasc Thorac Surg 2009;9 (2): 318–322.

5. Murkin JM, Adams SJ, Novick RJ, et al. Monitoringbrain oxygen saturation during coronary bypass sur-gery: a randomized, prospective study. Anesth Analg2007; 104 (1): 51–58.

6. Slater JP, Guarino T, Stack J, et al. Cerebral oxygendesaturation predicts cognitive decline and longerhospital stay after cardiac surgery. Ann Thorac Surg2009; 87 (1): 36–44; discussion 44–35.

7. Fischer GW, Benni PB, Lin HM, et al. Mathematicalmodel for describing cerebral oxygen desaturation inpatients undergoing deep hypothermic circulatoryarrest. Br J Anaesth 2010; 104 (1): 59–66.

32. FOUNDATION OF TISSUE OXYGENATION: OPTIMIZINGSYSTEMIC BLOOD FLOW BY TRANS-OESOPHAGEAL DOPPLER(TED) MONITORING

Schober P., Loer S. A., Schwarte L. A.

VU University Medical Center Amsterdam, Department of

Anaesthesiology

Introduction: Systemic blood flow is a key link in the chainof oxygen transfer from alveoli to mitochondria and istherefore one of the main determinants of tissue oxygen-ation. Impaired systemic blood flow is known to readilycompromise oxygen delivery and hence, maintenance ofan adequate cardiac output (CO) is considered pivotal inmaintaining an adequate tissue oxygenation. Clinically,thermodilution techniques, which require insertion of apulmonary artery catheter, are considered the goldenstandard for determinations of CO. Major risks, high costsand a considerable time needed for PAC insertion usuallylimit the use of this technique to patients with high risk ofhaemodynamic instability. Alternative techniques such aspulse contour analysis or transoesophageal echocardiog-raphy are also either invasive or require expert operators,so that CO is most often not assessed at all in patients at riskof developing tissue hypoxia. In these patients, transo-esophageal Doppler (TED) ultrasound of the descendingaorta may offer a promising alternative. The technique isminimally invasive, is easy to learn, allows quick andcontinuous assessment of CO and does not require a sterileenvironment, making TED technology suitable for rou-tine clinical use.

Technical principles: The TED probe is inserted into theoesophagus similar to a gastric tube and subsequentlyaligned with the descending aorta. Emitted ultrasoundwaves (typically with frequencies of 4 to 5 MHz) arescattered by erythrocytes travelling in the ultrasound beamand partially reflected back to the probe. A change in thefrequency of waves (in this case ultrasound waves)reflected by moving objects (here: erythrocytes) is knownas Doppler effect, and the shift between emitted andperceived frequency is directly proportional to velocity ofthe moving object. This principle allows to calculate

38 Journal of Clinical Monitoring and Computing

descending aortic blood flow velocity and to plot bloodflow velocity (cm sec-1) against time (sec). The areaunder the systolic portion of this curve (cm sec-1Æsec)corresponds to the distance (cm) the blood column hasmoved forward in the aorta during systole. Multiplyingthis distance (cm) with aortic cross sectional area (cm2)yields descending aortic stroke volume (cm3), from whichsystemic stroke volume and hence CO can be estimatedassuming a constant distribution of blood flow betweenthe descending aorta (� 70%) and the coronary andbrachiocephalic arteries (� 30%).

Limitations and clinical validity: Note that we basicallyhave two unknowns in the computation of CO: aorticdiameter and distribution of blood flow. Depending on thedevice, aortic diameter is either estimated with integratedM-mode ultrasound, or systemic stroke volume is calcu-lated using a validated nomogram based on patient’s age,height and weight. Technical limitations mainly derivefrom the assumptions needed to translate blood flowvelocity to CO and may prevent the technique frommeasuring exact absolute CO values. Indeed, variousstudies suggest that, while TED does not systemicallyunder- or overestimate CO, individual CO values canconsiderably differ from values obtained with referencemethods. However, trend monitoring should generally bepossible despite these limitations as long as the basic con-ditions, i.e. aortic diameter and blood flow distribution,remain constant. This has been confirmed by numerousstudies which have shown that TED reliably follows COchanges over time. Therefore, while TED may not be anideal technique when exact absolute values of CO arerequired, it is useful especially as a non-invasive and con-tinuous technique for trend monitoring of CO. However,it is important to realize that changes of the basic condi-tion, e.g. sudden changes of aortic diameter and bloodflow distribution due to acute hemorrhage, may likely leadto inconsistent or even misleading CO readings.

Hemodynamic optimization: In clinical practice, non-inva-sive monitoring of CO changes may be useful for earlydetection and management of hemodynamic deteriora-tion. Herein, TED also aids the clinician by measuringtime-related and accelerometric parameters that allowindirect assessment of preload, afterload and myocardialcontractility, which are however beyond the scope ofthis abstract. In addition, TED can be used to optimizeCO in patients without obviously impaired systemicblood flow. Herein, especially the role of TED tooptimize fluid load has been well defined in the litera-ture. Since the normal, not hypervolaemic hart operateson the ascending limb of the Starling curve, a fluid boluswill result in an increased stroke volume. In contrast,absence of an adequate increase suggests that the hartoperates on the flat portion of the Starling curve and that

further filling will result in volume overload. In thiscontext, an increase is generally considered adequate if itexceeds 10% following a colloid bolus of 3 ml kg-1

bodyweight. Thus, using TED together with the Starlingprinciple allows to optimize preload while avoidinghypervolaemia. A total of 10 studies using TED guidedvolume management strategies in perioperative andtrauma patients conclusively report beneficial effects inthe Doppler-guided groups. TED managed patientsrequired fewer days on an intensive care unit and wereearlier medically fit for discharge from hospital. More-over, these studies demonstrate that Doppler-guided fluidreplacement reduces the risk of postoperative complica-tions and morbidity. Reductions in the incidence ofpostoperative nausea as well as a shorter time to recoveryof gut function and resumption of enteral nutrition havealso been reported in the volume-optimized groups.While none of the studies have actually investigated theinfluence of TED guided volume management on tissueoxygenation, it is likely that the observed improvementin outcome is ultimately due to improvements in tissuemicrocirculation and oxygenation. In contrast to the welldescribed beneficial effects of TED guided volumemanagement in perioperative patients, studies assessingthe ability of TED to guide hemodynamic managementin other patient populations or with inotropic andvasoactive drugs are lacking.

Conclusions: Transoesophageal Doppler measurementsof descending aortic blood flow velocity is a minimallyinvasive, easy to learn and quick method for continuoustrend monitoring of CO. Various studies have demon-strated a reduced postoperative morbidity and shorterlength of hospital stay in perioperative patients managedwith TED, however further studies are needed to deter-mine the role of TED in guiding inotropic and vasoactivetherapies and to asses the influence of TED guidedhemodynamic management on tissue oxygenation.

REFERENCE

An extended list of references can be obtained from theauthor.

33. HUMAN FACTORS IN THE OR

Jan Maarten Schraagen1, Ton Schouten2,

Meike Smit4, Felix Haas5, Dolf van der Beek4,

Josine van de Ven, Paul Barach2,3

1TNO Human Factors, Soesterberg, The Netherlands, 2 Depart-

ment of Perioperative Care and Emergency Medicine, University

Abstracts 39

Medical Center Utrecht, 3 Patient Safety Center of the University

Medical Center Utrecht, The Netherlands, 4TNO Quality of Life,

Leiden/Hoofddorp, The Netherlands, 5Department of Pediatric

Cardiothoracic Surgery, University Medical Center Utrecht, The

Netherlands

Objective: Pediatric cardiac surgery (PCS) has a low errortolerance and demands high levels of cognitive and technicalperformance. The risk of technical failure during operations isrecognised, but there is evidence that further improvements insafety depend on systems factors, in particular, effective teamskills. The hypotheses that small intraoperative non-routineevents (NRE’s) can escalate to more serious situations and thateffective teamwork can prevent the development of serioussituations, were examined to develop a method to assess theseskills and to provide evidence for improvements in trainingand systems.

Methods: This mixed-method design, using bothquantitative and qualitative measures, used trained humanfactors observers that observed and coded NRE’s andteamwork elements from the time of patient arrival intothe operating room (OR) to the patient handover in theintensive care unit. Real-time teamwork observationswere coupled with questionnaires on safety culture,microsystems preparedness measures, operative duration,assessed difficulty of the operation and patient outcomemeasures. Behaviour was rated whether it hindered orenhanced teamwork.

Results: 40 PCS cases were observed. Surgeons displayedbetter teamwork during complicated procedures, partic-ularly during the surgical bypass/repair epoch. Moreprocedural non-routine events were associated with amore complicated post-operative course (Muncomplicated =9.08; Mminor complications = 11.11; Mmajor morbidity = 14.60,F(2,26) = 3.46, p < .05). A rapid 5 question questionnairefilled out immediately after the operation correlatedsignificantly with the complexity (r = .54), duration(r = .62), number of non-routine events (r = .31) andpatient outcome (r = .56). The total number of teambehaviours was associated with case complexity (r = .39),but this association disappeared when duration of theoperation was taken into account (r = .04). Proceduralnon-routine events decreased substantially over time(M1 = 13.5; M2 = 7.1, F(1,37) = 33.07, p < .001).

Conclusions: Structured observation of effective team-work in the operating room can identify substantivedeficiencies in the system, even in otherwise successfuloperation. High performing teams display good team-work when operations become more difficult. Betterteamwork does not directly lead to better patient out-comes.

34. REFLECTANCE SPECTROPHOTOMETRY

Lothar A. Schwarte, Stephan A. Loer, Patrick

Schober

Department of Anaesthesiology, VU university medical center,

Amsterdam, The Netherlands

Introduction & methods: The microcirculation is the essen-tial vascular compartment where blood-transported O2 isfinally delivered to the O2-consuming cells. Therefore it isof utmost importance that microcirculatory oxygenationis adequate to balance loco-regional metabolic demands.

A non-invasive, optical technique to measure micro-circulatory oxygenation is reflectance spectroscopy (alsocalled remission spectroscopy or -photometry). In brief,beams from a light source (e.g., from a high pressure xenonlamp) are guided to the tissue of interest (e.g., via a flexibleglass fibre cable), this light interacts with the illuminatedtissue and a fraction of the light is ultimately re-collectedand sent back via the light-guide to the analysing monitorsystem. The interaction of light within the tissue of interestincludes wavelength-dependent absorption at haemoglo-bin molecules, which depends on the degree of haemo-globin oxygenation, allowing calculation of themicrocirculatory haemoglobin O2-saturation (lHbO2)from the collected spectra. Herein the microcirculatoryoxygenation measured resembles a composite oxygenationmarker from microcirculatory arterioles, capillaries andvenoles. It thus reflects the loco-regional O2-balance be-tween O2-delivery and O2-consumption, and thus mi-crocirculatory O2-availability. Consequently, increasedO2-delivery (e.g., increased microcirculatory perfusion) ordecreased O2-consumption (e.g., by hypothermia) wouldincrease lHbO2, whereas decreased O2-delivery (e.g.,hypoperfusion) or increased O2-consumption (e.g., hyper-metabolism) would decrease lHbO2.

Advantages of reflection spectrophotometry: Reflectancespectroscopy is a non-invasive optical technique, notrequiring any toxic dyes (like Pd-porphyrine techniques)or traumatic instrumentation (like Clark-type PO2-needleelectrodes or microdialysis catheters). Thus, spectroscopyappears advantageous in bench-to-bedside research,because this method can be applied both in animalexperiments and also in patients. In addition, the mea-surement may be classified as continuous, whereas the tra-ditional technique of tonometry, particularly salinetonometry, require prolonged equilibration times.

In current systems, the technique of reflection spec-troscopy may be combined with techniques of laserDoppler flowmetry (LDF), allowing the simultaneous

40 Journal of Clinical Monitoring and Computing

assessment of microcirculatory oxygenation and micro-circulatory blood flow.

Which spot to measure? Given the heterogeneity of themicrocirculation between and within organs, one has topredefine a meaningful spot to measure microcirculatoryoxygenation. Obviously, to monitor microcirculatoryoxygenation during (or after) certain localized surgicalprocedures, the measurement spot is defined by the sur-gical procedure itself: After transplantation of a freemusculo-cutaneous flap, monitoring the microcirculatoryoxygenation of skin of the transplant allows detection ofarteriolar inflow reduction and/or venous outflow prob-lems, leading to venous congestion of the transplant.Thus, function of the non-accessible arteriolar and venousanastomoses can be deduced from a non-invasive surfacemeasurement. In contrast to the routine assessment usingclinical signs (skin colour, skin temperature and tissueswelling), this technique is more objective, better to bequantified and allows for a continuous measurements.

However, in the disciplines of anaesthesia, critical careand emergency medicine, it is often not a single organwhere oxygenation is endangered, but the entire body(e.g., by cardio-circulatory distress, including shock).Thus, the selection of the optimal measurement spot isnot as obvious as exemplified for surgical procedures.Traditionally it is argued, that focusing the monitoring onso-called vital organs (e.g., heart or brain) is advantageousherein, since hypoxygenation particularly of these organswould lead to disastrous consequences for patient out-come. Although the latter statement per se is obviouslycorrect, it misses the crucial aspect that the so-called vitalorgans are well protected from circulatory distress andreact relatively late in the sequence of circulatory distress.Reasons for this rather robust oxygenation of the so-calledvital organs are a well developed local autoregulation ofperfusion and a preferential perfusion in states of circula-tory centralization.

Thus, we argue that monitoring of organs or tissuesexplicitly not belonging to the so-called vital organsappears more advantageous, because these remote organswould react more early in states of circulatory distress andthus could serve as sentinel organs. Following this argu-mentation, this early detection of disturbances of micro-circulatory oxygenation in remote organs supports rapidsuspicion, rapid diagnosis and ideally rapid correction ofthe patient condition, even markedly before signs of tissuehypoxygenation occur in the so-called vital organs. Thiswould allow the switch from a reactive concept of therapyto a more proactive therapy concept of tissue oxygenation.This is fully in line with concepts of the golden hour ofshock, nominating the factor of time and timing as crucialin the setting of acute care.

The splanchnic region possesses several items thatrender it an excellent candidate as measurement spot formicrocirculatory oxygenation in this setting. As non-vitalorgans, the splanchnic organs participate strongly in theprocess of circulatory centralisation, e.g., via intensesplanchnic vasoconstriction. This response occurs early inthe sequence of cardio-circulatory distress, e.g., haemor-rhage or other forms of hypovolaemia. Within thesplanchnic region, the gastrointestinal tract is principallyaccessible via natural orifices, e.g., orally or rectally. Theoral route has been established for gastric tonometry, atraditional method to assess the adequacy of regionalperfusion/oxygenation. Although the technique oftonometry is hampered with multiple methodologicalproblems, the gastric mucosa remains an attractive mea-surement site for microcirculatory oxygenation, e.g., asnon-traumatic accessible part of the splanchnic region.

Results & discussion: As discussed, we selected thegastric mucosa as preferred measurement spot for most ofour studies so far. Since the impact of anaesthesia rele-vant pathologies (e.g., anaemia or haemorrhage) andanaesthesiological interventions on the microcirculatorygastromucosal oxygenation were unclear, we performeda series of studies on this subject, both experimental andclinical. For brevity of this abstract, the general keyfindings from our studies may be summarized as follows:

1. In various (models of) pathologies, microcirculatoryoxygenation may react earlier than systemic markers.

2. Even the direction of changes may differ between thesystemic circulation and the regional microcirculation.

3. Therapeutic options improving the systemic circulationdo not necessarily improve regional microcirculation.

4. Therapeutic options exist, that (so far in animalexperiments) selectively improve gastromucosalmicrocirculatory oxygenation, i.e., without markedimpact on systemic circulation.

A major task within anaesthesia, critical care andemergency medicine is to ensure adequate oxygenation ofthe patient. Consequently, we should extend care fromensuring blood oxygenation (as routinely monitored bystandard pulse oxymetry) to ultimately ensuring cellularoxygenation. This task can be supported by using tech-niques to more directly measure tissue oxygenation, e.g.,by reflection spectrophotometry. Besides technical pro-gress (e.g., need for smaller devices applicable in clinicalpractice) a number of principal questions still remain 5unsolved, e.g., addressing critical or goal values ofmicrocirculatory oxygenation. Further, the therapiesaiming at improvement of the microcirculation are farfrom established and thus demand further research.

Abstracts 41

REFERENCE

An extended list of references can be obtained from theauthors, via: L.Schwarte@vumc.nl.

35. SIMULATION IN HEALTHCARE: THE TECHNOLOGICALPERSPECTIVE

Eric Stricker, M.Sc., Dr. med. Marcus Rall

Center for Patient Safety and Simulation TuPASS, University

Hospital Tuebingen, Germany

No Astronaut would leave the planet without hundreds ofhours spend in the simulator to train all tasks he needs forhis mission. No Pilot would get his license withoutpassing a lot of simulated scenarios. No bridge would bebuilt without having simulated all thinkable influences.Simulation is used in many domains and in many differentways to minimize all risks or to prepare all professionals inthe best way. In all high risk industries, such as military,aviation, space flight, and nuclear power industries, thistechnique is well known and used very intensive. To get ahighly reliable system many tests, proofs and checks are anabsolute must. Is this the case in healthcare?

Simulation refers to the artificial replication of sufficientelements of a real-world domain to achieve a stated goaland it can help users understand and model real life sys-tems. Simulation should be used when it is expensive ordangerous to run the real systems. We can gain betterunderstanding of a system and identify problems. More-over we can test the potential effects of changes. Mosttechnical disciplines use simulation to optimize productdevelopment processes. Planning and development ismuch easier and less expensive with using differentmethods of simulation. A very common and useful tool isfor example the finite elements method (FEM). The finiteelement method is a numerical technique to findapproximate solutions in many mechanical engineeringdisciplines. FEM is nowadays a pure computer simulation.

Simulation has many faces, her some examples to get anoverview:

• Mathematical simulation is the formal modeling ofsystems has been via a mathematical model, whichattempts to find analytical solutions to problems whichenables the prediction of the behavior of the systemfrom a set of parameters and initial conditions

• Computer simulation has become a useful part ofmodeling many natural systems in physics, chemistryand biology, and human systems in economics andsocial science

• Physical simulation substitutes the real thing bysimilar or physically comparable objects; most of themare smaller models of the real thing.

• Interactive simulation or ‘‘Human-In-The-Loop

Simulation’’ means the human interaction with aSystem, or Humans as part of a physical simulation.Flight Simulators are probably the most well-knownrepresentatives of this type of simulation, the PatientSimulation is one too.

It is more and more common to hear simulations ofmany kinds referred to as synthetic environments or vir-tual reality. This label has been adopted to broaden thedefinition of simulation. The big field of virtual reality isgetting a more and more important part in the simulationfield or as part of interactive simulation.

Simulation in healthcare: Use of part task trainers is a veryold method to train healthcare professionals or medicalstudents. In the 1700 s Madame du Coudray created ‘‘themachine’’ to train midwifes in the court of King Louis XV.In healthcare we know a lot of computer simulation tools,so called medical microsimulators. They were developed tosatisfy the medical student’s needs to focus on the concep-tual understanding of medical procedures and to train themto identify and understand medical cases and the treatment.To train procedures, tasks or skills the so called basic sim-ulators find their application. Examples for basic simulatorsare resuscitation manikins to train chest compressions.There is a big variety of part task trainers for nearly everymedical skill available. Beside all skill trainer and low fidelitytraining manikins there are just a few high fidelity simulatorsavailable. The three big manufacturers follow two differentconcepts. Model driven simulators and free ‘‘on the fly’’programmable simulators. There is still an ongoing dis-cussion which concept is the best. From the technicalperspective the model driven simulators are real computersimulation with a manikin as interface. But if we have amore psychological look on it, simulation is not even themanikin and it inputs, outputs and mathematical behavior.It is more the whole training room in which the real cli-nician has to treat a virtual patient. David Gaba wrote in2004 ‘‘simulation is a technique, not a technology’’.

Discussion: How can clinicians experience the difficultiesof patient care without putting patients at undue risk? Howcan we assess the abilities of clinicians as individuals andteams when each patient is unique? These are questions thathave challenged medicine for years. In recent years, theseand related questions have begun to be answered in healthcare by the application of approaches new to medicine, butborrowed from years of successful service in other industriesfacing similar problems. We should concentrate on whichtool fits best to our educational needs. The big question is,how can technology help simulation instructors to fulfill all

42 Journal of Clinical Monitoring and Computing

educational needs? Some goals can be achieved with min-imal fidelity, others require very high fidelity. Some degreeof simulator and simulation fidelity is required to engageparticipants in a learning or evaluation activity. It needs aphysical fidelity, a conceptual an d an emotional fidelity. Ithas to look, feel and behave like a real patient, and it shouldhave the ability to draw the participant into the situation. Inone word the simulation and the instructor has to create arelevant experience for Simulation participants.

REFERENCES

1. Gaba D M; Qual Saf Health Care 2004;13(Suppl 1):i2–i10. doi: 10.1136/qshc.2004.009878

2. Gardner R, Raemer D,; SIMULATION INOBSTETRICS AND GYNECOLOGY; ObstetGynecol Clin N Am 35 (2008) 97–127

3. Good M, Gravenstein J. Anesthesia simulators andtraining devices. Int Anesthesiol Clin 1989;27: 161–6.

4. Gaba DM, DeAnda A.Acomprehensive anesthesia sim-ulator environment: re-creating the operating room forresearch and training. Anesthesiology 1988;69: 387–94.

5. Rall, M., & Gaba, D. M. (2005). Patient simulators. InR. D. Miller (Ed.), Anaesthesia (pp. 3073–3103). NewYork: Elsevier.

Weblinks

http: //www.symphonytech.com/articles/pdfs/simulation.pdfhttp: //en.wikipedia.org/wiki/Simulationhttp: //en.wikipedia.org/wiki/Finite_element_method

36. PROBABILISTIC APPROACHES TO SOLVING THE INVERSEPROBLEM OF STATE AND PARAMETER ESTIMATION FORMECHANISTIC MODELS OF PHYSIOLOGY: FROM THEORYTO PRACTICE

Sven Zenker

Dpt. of Anaesthesiology & Intensive Care Medicine, University of

Bonn Medical Center, Bonn, Germany

The inverse problem of parameter and state estimationfrom available observations for mechanistic mathematicalmodels of physiological processes is one of the mostchallenging steps in moving from theory to application.This talk will, using practical examples, present an over-view of established as well as recently developed numer-ical approaches that approximate full posteriordistributions on joint parameter and state space, poten-tially enabling meaningful inference even in challengingsituations where traditional point estimators encounterdifficulties.

Abstracts 43