Journal of Human Hypertension (1997) 11, 163 169 1997 Stockton Press. All rights reserved 0950-9240/97 $12.00 Evaluation of the Critikon 8100 and Spacelabs 90207 non-invasive blood pressure monitors using a test simulator Department of Medical Physics and Medical Engineering, Royal Infirmary of Edinburgh, Edinburgh, UK The Critikon Dinamap 8100 and the Spacelabs 90207 consistency than the Dinamap which showed a higher ambulatory non-invasive blood pressure (NIBP) moni- standard deviation under all the conditions. The relators were evaluated using a test simulator using an tively high standard deviation of the recordings made evaluation protocol which covered a wide range of by the Dinamap could explain the non-systematic errors simulated pressures (with five determinations at each of found in some evaluations. Both instruments recorded six steps from 60/30 to 200/150 mm Hg), pulse rates pressures within 5 mm Hg of the target over the range (from 40 to 200 bpm), artefact levels (simulated motion of pressures and pulse rates, and coped well under conand tremor artefact) and pulse strengths (down to 10% ditions of severe artefact and weak pulsations, either by of the nominal strength). Determinations were made at signalling inability to record or by recording satisfac- 5 min intervals. The average and standard deviation of torily. the five measurements at each condition were calculated. The Spacelabs recorded pressures with a greater Keywords: NIBP monitors; evaluation; simulators Introduction Oscillometric non-invasive blood pressure (NIBP) measuring instruments are in widespread use in health care, with a significant growth in the use of these devices in hospitals in recent years; in our teaching hospital their number, per 100 hospital beds, increased over five-fold from about three in early 1987 to 17 in early 1996. While several tech- niques may be used to measure blood pressure (BP), 1 the oscillometric technique is most commonly used for automated NIBP measurements, being found in over 80% of instruments. 2 Oscillometric NIBP monitors record the systolic, mean and diastolic pressures by relating the cuff pressure to the shape of the waveform of low ampli- tude ( 2 mm Hg) pressure oscillations. The oscil- lations are generated in a cuff placed around a limb as its pressure is reduced from above the systolic to below the diastolic pressure. The mean BP is estimated as the cuff pressure corresponding to the maximum amplitude of the oscillometric wave- form. 3 The diastolic and systolic pressures are empirically determined based on the relationship between the cuff pressures corresponding to specific fractions of the peak height of the envelope of the oscillations. Different manufacturers use different algorithms to interpret the oscillometric waveform, 4 Address for correspondence: Dr John Amoore, Department of Medical Physics and Medical Engineering, Royal Infirmary of Edinburgh, Lauriston Place, Edinburgh EH3 9YW, UK Received 30 January 1996; revised 2 January 1997; accepted 6 January 1997 and their own proprietary signal processing techniques to improve the ability of the monitor to record the BP under conditions of low pulse strength and movement artefact. The consequence of the device-specific signal pro- cessing and algorithm is that each different NIBP monitor has to be validated to check if it correctly interprets the BP. Evaluation standards and protocols have been drawn up, 5 7 which compare the NIBP monitor with direct intra-arterial BP readings or with manual auscultatory measurements. 8 10 These clinical trials evaluate the monitor over a wide range of BPs, with multiple readings at each pressure. This is of necessity a costly business requiring the recruitment of many patients. To reduce the need for trained observers a technique has been developed which involves recording on video-tape the measurement process (mercury sphygmomanometer together with the auscultatory signal) using a camcorder. 11 This technique pre- serves the advantages of evaluating an NIBP monitor against the standard auscultatory method and by recording the process allows the measurements to be made and checked later. The development of NIBP test simulators presents the opportunity of developing alternative protocols for evaluating NIBP monitors. 7,12,13 These protocols can supplement, but not replace, clinical trials. We have developed an evaluation protocol using test simulators which has been described elsewhere. 13 Here we show how this protocol can be used to help evaluate two NIBP monitors, providing functional information not readily obtainable during conventional clinical trials.
164 Materials and methods affects the measurement of the pressure. 14 We recorded the pressures over a range of pulse rates A number of NIBP simulators, which generate oscillometric from 40 bpm to 200 bpm with the simulated presscommercially pulses in response to cuff inflation, are ure kept fixed at 120/80 (93) mm Hg. Most NIBP available. 7 Waveforms simulating a monitors adjust the rate of change of cuff pressure wide range of pressures and pulse rates may be gen- with the pulse rate to ensure accurate determinations erated, with their repeatability specified to within 1 at low pulse rates, while taking advantage mm Hg. The test simulators enable NIBP monitors of the higher sampling frequency at higher pulse to be tested over a wide variety of conditions such rates to reduce the determination time. We recorded as may be encountered in clinical practice. The the time for each determination. Five determinations simulators also record a number of parameters at each pulse rate were made. describing the determination, such as peak inflation The BP-Pump can be used to test the ability of the pressure, determination time and deflation rate. The NIBP monitor to cope with artefact. A low frequency Bio-Tek BP Pump (Bio-Tek Instruments Inc, High- Motion artefact may be added to the oscillometric land Park, Box 998, Winooski, Vermont 05404 waveform in various degrees of severity from 1 to 2 0998, USA) was used in this study. to 5 and up to 10. (Level 1 artefact is simulated by The cuff hose of the monitor under test was connected adding noise with a peak-to-peak amplitude of to the test simulator which incorporates an 0.2 ml in relation to the signal amplitude of 1.2 ml. internal cuff, designed to simulate the volume and Level 2 is twice the noise amplitude. We tested the compliance of an adult cuff wrapped around a limb. monitors up to noise levels 5, that is noise amplitude The simulator was controlled by a personal computer of 1 ml.) Similar ranges of severity of mixed which recorded the peak inflation pressure low and high frequency Tremor artefact were added. and the determination time of the NIBP monitor as In all cases the simulator was set to generate a press- determined by the simulator. 13 The Critikon 8100 ure of 120/80 mg Hg at a pulse rate of 80 bpm. The was also directly connected to the personal computer average Bias and standard deviation of five success- via its serial link, and each measured pressure ive determinations (at 5-min intervals) were determ- was recorded. The Spacelabs stores its recorded ined at each artefact setting. measurements in its internal memory. Its measure- The ability of a monitor to cope with weak pulsations ments were transferred after the evaluation to a spreadsheet was tested by reducing the amplitude of the and linked to the data from the simulator oscillometric waveform down to 10% of the nominal for analysis. typical amplitude. The Bio-Tek BP-Pump califor We compared the difference (Bias) between the brates its nominal 100% amplitude as a 1.2 ml volume pressure set on the simulator (the measurement displacement. This corresponds to the typical standard for the analysis) and the pressure recorded amplitude of the oscillometric waveform of 2 by the simulator. Even when presented with a fixed mm Hg. simulated waveform NIBP recordings will vary from We demonstrate the role that NIBP test simulators determination to determination. 12,14 Hence, each can play as a complementary tool for the evaluation simulated pressure waveform was repeated five of NIBP monitors by assessing the Critikon Dinamap times (with an interval of 5 min between each 8100 NIBP monitor and the Spacelabs 90207 ambu- determination). The averaged Bias was calculated as latory NIBP recorder. The Dinamap 8100 is in widespread was the sample standard deviation. use in health care facilities, and has been The evaluation protocol covered a range of simu- evaluated according to the British Hypertension lated pressures, pulse rates, weak pulsations, and Society guidelines, 9 as has the Spacelabs ambulat- motion and tremor artefact (Table 1). The first part ory monitor. 8 of the protocol tested the monitor at six different pressures ranging from low (60/30 mm Hg) to high (200/100 mm Hg), with five determinations at each Results target pressure. The bias and standard deviation Both NIBP instruments were tested over a range of were calculated at each pressure. We recorded the pressures from 60/30 to 200/150 mm Hg with the peak cuff inflation pressure and calculated the dif- pulse rate fixed at 80 bpm. The average of the difference ference between it and the previous systolic pressure, between the measured pressure and the set ignoring the first determination at each pressure simulator pressure for each of the five determinations step. This difference reflects the ability of the monitor at each pressure (bias) and the standard to adjust its cuff inflation pressure to the pre- deviation were calculated (Figure 1). The standard vious systolic pressure. deviation of the Critikon was greater than the Space- We then presented the monitor with a ramped labs. pressure sequence to test the ability of the monitor We calculated the difference between the peak to follow changing pressure patterns. The pressure cuff inflation pressure and the previous systolic setting of the simulator was ramped up from 60/30 pressure at each of the pressure steps (Figure 2); mm Hg to 255/195 mm Hg and then down to 60/30 both instruments adjusted their peak cuff inflation mm Hg with the sequence repeated. The differences pressure based on the previous systolic. between the measured pressures and the target Both instruments were able to follow a ramped simulator pressures were recorded. pressure changed with the Bias keeping within ±10 The cuff pressure is sampled by the oscillometric mm Hg over the systolic pressure range from 60 255 waveform at the pulse rate, and hence the pulse rate mm Hg (Figure 3). While we tested both units at a
Table 1 NIBP monitor evaluation protocol 165 Test description Simulated pressure Pulse rate Pulse strength (%) Pressure range Tests at six simulated pressures: Fixed pulse rate Fixed at 100% 60/30 (40) 120/80 (93) 80 bpm 80/50 (60) 150/100 (116) 100/65 (76) 200/150 (166) Ramp Pressure ramps from 60/30 to Fixed pulse rate Fixed at 100% (single determination at 255/195 to 60/30 80 bpm each step) Pulse rate Fixed pressure Pulse rates: 40, 60, Fixed at 100% 120/80 (93) 80, 120, 160, 200 Artefact Fixed pressure Fixed pulse rate Fixed at 100% Motion: Low frequency 120/80 (93) 80 bpm noise Tremor: Mixed low and high frequency noise Pulse strength Fixed pressure Fixed pulse rate Pulse strength: 100%, 120/80 (93) 80 bpm 75%, 50%, 25%, 10% Note: Unless indicated otherwise, five successive determinations are made at each setting within each test group. All pressures are in mm Hg. systolic of 255 mm Hg it should be noted that they 60/30 to 200/150 mm Hg. These results suggest that are only specified up to systolic pressures of 250 the relatively high variability of the determinations mm Hg. However, the high systolic pressure was by the Dinamap may account for the non-systematic included in order to establish whether they could errors which caused it to fail the BHS society evalu- successfully record a lower pressure (150 mm Hg) ation protocol. 9 In contrast, the Spacelabs produced after the peak high pressure. consistent measurements, in agreement with earl- Figure 4 shows the pressures recorded by both ier validations. 8 monitors over 25 successive determinations, five at The assessment of an NIBP monitor should each of five pulse rate over the range from 40 include an evaluation of its ability to record accurately 200 bpm. The determination time increased with the patient s BP over a wide range of con- decreases in pulse rate (Figure 5) reflecting the ditions including pressures and pulse rate. Conventionally reduction in deflation rate at the lower pulse rates the assessment is made by testing the in order to maintain consistent measurements. monitor over a range of patients and comparing the Figure 6 shows the Bias and Error of the recorded monitored pressure to the patient s actual pressure systolic pressure from two monitors when presented as determined by an arterial line, or by a gold standard with increasing levels of Motion and Tremor artefact. technique such as the auscultatory method. The target pressure was 120/80 (93) with a NIBP simulators provide an alternative assessment pulse rate of 80 bpm. Figure 6 shows only the systolic tool, but instead of comparing the monitor s deter- pressures; the effects of the artefact on the dias- mination to the actual patient pressure, the moni- tolic and mean pressures were similar. The Critikon tor s determination is compared to the notional notified the operator of unacceptably high levels of pressure of a simulated patient waveform. Figure 1 artefact at Tremor Level 5 (ee error code); the Spacelabs shows the difference between the pressures managed to record pressures at this high arte- recorded by the Critikon 8100 and the Spacelabs fact level, albeit with an increased Bias and variability. 90207 and the notional pressure of the waveform simulated by the test instrument. Clearly therefore, Both monitors successfully recorded pressures the extent to which the Bias shown in Figure 1 when the pulse strength was reduced to 25% of the matches that determined from clinical trials will nominal 2 mm Hg oscillometric waveform (Figure depend on the extent to which the simulated wave- 7), but neither could record pressures at 10% form matches the waveform of the patient. Nonetheless, pulse strength. the simulated waveform is consistent, and hence should enable a comparative assessment of the various monitors to be made. Discussion The average of the measurements at each simulated We tested both the Critikon Dinamap 8100 and the condition recorded by the Dinamap and the Spacelabs 90107 over a wide range of conditions. Spacelabs were within 5 mm Hg of each other. The main finding was the greater consistency of the O Brien et al 9 showed a non-systematic systolic measurements made by the Spacelabs over all the error of 1 ± 7 mm Hg for the Dinamap. The slightly conditions, with comparatively large determination lower standard deviation found in this study may to determination variations recorded by the Dina- reflect the contribution of physiological variations to map. The average bias of each instrument was however the variability observed in the clinical trial. How- within 5 mm Hg of the simulator target and of ever, this study did not find the systematic error in each other over the full range of pressures from the diastolic measurements suggested by the 6 ± 7
166 Figure 1 The Bias and standard deviation of five successive pressures recorded by the Critikon Dinamap 8100 and the Spacelabs 90207 at each of six pressures covering the range of pressures from 60/30 mm Hg to 200/150 mm Hg. The X axis shows the pressure (in mm Hg) at each simulation. shown how the test simulator correctly interpreted the effects of the different algorithms included in the Nellcor N3100 monitor. 16 It provides two algorithms to accommodate different interpretations of non- invasive BP measurements between Japan and Europe. The adjusted option was included for the Eur- opean market and involves a 2 mm Hg decrease in the systolic pressure and a 5 mm Hg increase in the diastolic pressure. The measurements recorded by the simulator reflected these differences. The simulator enabled us to test the ability of the monitor to follow large abrupt changes in BP, by presenting the ramped change in pressure. This test is useful, as most NIBP monitors base their cuff mm Hg observed in the clinical trial. We do not have an explanation for this, other than to note that another study recorded a non-systematic diastolic error of 0.7 ± 11 mm Hg. 15 Certainly, the results suggests relatively large non-systematic errors in the recordings of the Dinamap. While the simulators are designed to generate realistic waveforms based on recordings from patients, no reports validating the accuracy of the waveforms have been reported. The waveforms are however repeatable, enabling the consistency of measurements made by NIBP monitors to be assessed and to evaluate the relative differences between NIBP monitors. 16 In earlier work we have
167 Figure 2 Pressure above the systolic pressure to which the cuff is inflated for Critikon 8100 and the Spacelabs 90207. Figure 4 The systolic, mean and diastolic pressures recorded by each monitor as the pulse rate was increased through the range 40, 60, 80, 120, 160 and 200 bpm. Five determinations were made at each pulse rate at intervals of 5 min. The simulated pressure was kept constant at 120/80 (93) mm Hg. inflation pressure according to the previously recorded systolic pressure. In unpublished work we found one monitor which was unable to accurately record low pressures following a high pressure. It remembered the previous high systolic reading and, presented with a lower pressure and not finding the expected high systolic pressure, appeared to increase the gain of its pressure transducer amplifier and consequently interpreted noise as the start of the oscillometric waveform. Controlled artefact can be simulated and the ability of the monitor to accurately record pressures during the presence of artefact can be assessed. This is important as it follows from the very nature of automated NIBP monitors in clinical use that medical staff will not necessarily be observing the patient Figure 3 The systolic pressures recorded by the Dinamap and the Spacelabs as it was presented with two ramped changes of press- ure. The X axis shows the simulated systolic pressure. while the pressure is being recorded. Hence the monitor could be recording pressures while the patient s limb is moving without staff being aware of the movement. A monitor which provides apparently accurate recordings of BP which are significantly influenced by artefact may tend to mislead medical staff. It is much better that the monitor identify in some way the presence of artefact than attempt to provide a reading which is in error. Both in this study and in others 17 an increased variability of the recordings in the presence of simulated artefact has been observed. In addition, we observed that high levels of artefact typically lead to higher readings in monitors which do not cope well with artefact (Figure 6). This is because the presence of artefact superimposed on the cuff pressure leads to the monitor detecting an apparent oscillometric waveform at a higher cuff pressure. The evaluation of the ability of an NIBP monitor to cope with artefact is perhaps particularly relevant for ambulatory instruments. 18 NIBP simulators provide an objective
168 Figure 5 The time taken to record pressures at each of six pulse rates from 60 200 bpm. The average time over five successive recordings was determined. Figure 7 Effects of reducing the pulse strength on the Bias and standard deviation of the recorded systolic pressure. The target pressure was 120 mm Hg. Besides simply comparing the pressure determinations of a particular monitor with standardised waveforms, NIBP simulators can provide other useful measures of performance. The ability of NIBP monitors to rapidly make the measurement and with minimal inflation of the cuff helps to reduce the stress to the patient caused by the cuff inflation. Monitors adjust their rate of cuff deflation (or cuff inflation if the measurement is made during inflation) to maintain reproducibility by ensuring that the changing cuff pressure is adequately sampled by the pressure oscillations at the pulse rate (Figure 5). Technological advances have enabled the determination time to be reduced to 20 30 s compared to the 40 60 s of the early generation of monitors such as the Critikon 845. 12 Equally important is the ability to record the pressures without undue inflation of the cuff. Figure Figure 6 Effects of increasing levels of Motion (M1, M2, and M5) 2 compares the pressure to which the cuff is inflated and Tremor (T1, T2, T5) artefact on the Bias and standard deviabove the systolic pressure for the two monitors. ation of the recorded systolic pressure. The target pressure was 120 mm Hg. (ee) indicates that the monitor displayed its inability Both monitors adjusted their peak cuff inflation to record because of high levels of artefact. pressure in relation to the previous systolic peak, with the Dinamap generating lower cuff pressures at method of studying the effects of artefact on the the very low systolic pressures. This may reflect its recordings of pressure by NIBP monitors. In this neonatal mode of operation; the Spacelabs is study the Spacelabs 90207 was shown to cope designed for adult only use. extremely well with severe levels of artefact. The We have presented an evaluation of NIBP monitors Bio Tek BP-Pump provides a range of severities of using commercially available test simulators. artefact. Data needs to be collected to determine the We believe that such devices can provide an objective severity of artefacts encountered in the clinical setting. and important and significant adjunct to the clinical evaluation of NIBP monitors. Test simulators The NIBP simulator can also evaluate how well a can assess the consistency of pressure monitor copes with weak pulsations. Again, data measurements, check the response to abrupt needs to be collected to evaluate the amplitude of changes in pressure, examine the ability to cope with artefact and low pulse strengths and record details of the determinations such as the determi- nation time and the peak cuff inflation pressure. By the oscillometric waveform in patients with weak pulsations in order to decide what levels of pulse strength an evaluation protocol should include.
enabling an NIBP monitor to be checked over a wide sphygmomanometers. Hypertension 1993; 21: 504 range of pressures and pulse rates they may enable 509. clinical trials to be simplified, perhaps reducing the 7 Ng K-G, Small CF. Review of methods and simulators number of patients required for each evaluation. A for evaluation of non-invasive blood pressure moni- tors. J Clin Eng 1992; 17: 469 479. further important potential use for the NIBP simu- 8 O Brien E, Mee F, Atkins N, O Malley K. Accuracy of lator will be to reassess the adequacy of NIBP monithe Spacelabs 90207 determined by the British Hypertors after prolonged periods of clinical use. tension Society protocol. J Hypertens 1991; 9: 573 574. Acknowledgements 9 O Brien E, Mee F, Atkins N, O Malley K. Short report: Accuracy of the Dinamap portable monitor, model 8100 determined by the British Hypertension Society The financial support of the Chief Scientists Organprotocol. J Hypertens 1993; 11: 761 763. isation of the Scottish Office Home and Health 10 Alpert BS. Validation of CAS model 9010 automated Department through the Advisory Panel on Evalu- blood pressure monitor: children/adult and neonatal ation of Medical and Scientific Equipment and studies. Blood Press Monit 1996; 1: 69 73. Health Service is gratefully acknowledged. We 11 O Brien E et al. A new audiovisual technique for recthank the reviewer for his comments and Dr P Pad- ording blood pressure in research: the Sphygmocorder. field of the Department of Medicine at the Western J Hypertens 1995; 13: 1734 1737. General Hospital in Edinburgh for helpful advice in 12 Amoore JN. Assessment of oscillometric non-invasive preparing the paper. blood pressure monitors using the Dynatech Nevada CuffLink analyser. J Med Eng Tech 1993; 17: 25 31. 13 Geake WB, Amoore JN, Scott DHT. An automated system for the functional evaluation of oscillometric non- References invasive blood pressure monitors. J Med Eng Tech 1 O Brien E, Fitzgerald D. The history of blood pressure 1995; 19: 162 176. measurement. J Hum Hypertens 1994; 8: 73 84. 14 Hatsel CP. Cardiac cycle phase uncertainty: another 2 Ng K-G, Small CF. Survey of automated non-invasive source of error in indirect blood pressure measureblood pressure monitors. J Clin Eng 1994; 19: 452 475. ment. J Med Eng Tech 1992; 16: 157 158. 3 Ramsey M III. Noninvasive automatic determination of 15 Ornstein S, Makert G, Litchfield I, Zemp L. Evaluation mean arterial blood pressure. Med Biolog Eng Comput of the DINAMAP blood pressure monitor in an ambulatory 1979; 17: 11 18. care setting. J Fam Pract 1988; 26: 517 521. 4 Sapinski A. Standard algorithm of blood-pressure 16 Amoore JN, Geake WB, Scott DHT. Oscillometric noninvasive measurement by the oscillometric method (Letter). blood pressure measurements: the influence Med Biolog Eng Comput 1994; 32: 599. of the make of instrument on the reading? Med Biolog 5 O Brien E et al. The British Hypertension Society protocol Eng Comp 1997; 35: 131 134. for the evaluation of automated and semi-auto- 17 Mieke S et al. Zur Me sicherheit nichtinvasiver oszil- mated blood pressure measuring devices with special lometrischer Blutdruckme geräte. Anaesthesist 1993; reference to ambulatory systems. J Hypertens 1990; 8: 42: 38 43. 607 619. 18 West JNW et al. Effect of unrestricted activity on accu- 6 White WB et al. National standard for measurement of racy of ambulatory blood pressure measurement. resting and ambulatory blood pressure with automated Hypertension 1991; 18: 593 597. 169