Anaesthesia, 1992, Volume 47, pages 693-696 APPARATUS
An automated system for testing the accuracy of patient-controlled analgesia devices
D. J . HAWKINS, D. R . D. ROBERTS, A . H. ILSLEY, J. L. PLUMMER
AND
H . OWEN
Summary A system was developed to test the accuracy of patient-controlled analgesia devices in situations simulating clinical use. Bolus requests are made automatically at predetermined intervals, and the infusate delivered is measured and recorded without the need for operator presence. To ensure clinical relevance, the bolus request times used in this study corresponded to a pattern typical of those requested by patients on the ward. Graseby, Abbott Provider 5.500 and W A C patient-controlled analgesia devices were tested and found to deliver reasonably accurately over a 24 h period. However, when an infusion was started in an unprimed system or a f e r a period of no bolus requests in a bolus-only mode the Graseby and W A C machines under-delivered. This system provides a means of testing patient-controlled analgesia devices operating in any delivery mode.
Key words Analgesia; patient-controlled. Equipment; infusion pumps, computer.
Over the last 20 years, patient-controlled analgesia (PCA) has developed from an investigative tool [I] to a well-used and desirable form of postoperative pain relief [2]. The choice of suitable PCA devices has increased accordingly although . . . ‘many are available for use without a single publication to document the safety and utility of the device in the routine patient care situation’ [3]. Where articles have been published reviewing these devices, most report their accuracy [3-61 assessed volumetrically, or more usually gravimetrically, from collection of infusate delivered. In these studies the output was measured manually, and continuous infusion and bolus doses were tested separately with no effort being made to simulate clinical use. Testing a large number of PCA devices under simulated clinical conditions would be very time consuming and to request bolus doses at timed intervals, with and without a background infusion, would be even more laborious. One solution is a computer-controlled system where bolus requests can be made by the computer and the weight of infusate delivered recorded automatically. The aim of this study was to develop a system for the automated assessment of PCA devices with particular attention to their accuracy in a simulation of clinical use. Three types of electronic PCA devices were studied using this system.
Methods A TIOOO lap-top computer (Toshiba corporation, Tokyo, Japan) was used to control a system which could initiate demands and gravimetrically measure the fluid delivered by the PCA devices. Narrow bore plastic tubing connected the PCA device to a long 14G needle which rested on the bottom of a measuring cylinder below a layer of light oil introduced to reduce inaccuracies due to evaporation. Fluid delivered by each PCA device was continuously weighed on a calibrated Mettler PM2000 electronic balance (Mettler Instruments, AG Zurich, Switzerland). The computer was programmed to interrogate the balance hourly; the weight and time were then automatically recorded on computer disk. The computer program simulated the bolus requests of a patient via a purpose-built interface box connected to the PCA device. When a bolus was requested the interface box activated the PCA device either directly by an electrical connection or by a pneumatic connection (for the Graseby PCAS). The pneumatic pressure was provided by a Maxima pneumatic aquarium pump (Ralph C. Hagen, Leeds, UK). Communication was via serial connections made between the computer and interface box and between the computer and the electronic
D.J. Hawkins, FRCAnaes, D.R.D. Roberts, FRCAnaes, Senior Registrars, A.H. Ilsley, PhD, J.L. Plummer, PhD, Principal Hospital Scientists, H. Owen, MD, FANZCA, Senior Staff Specialist, Department of Anaesthesia and Intensive Care, Flinders Medical Centre, Bedford Park, South Australia 5042, Australia. Correspondence should be addressed to Dr H. Owen. Accepted 23 December 1991. 0003-2409/92/080693 + 04 $03.00/0
@ 1992 The Association of Anaesthetists of Gt Britain and Ireland
693
694
D.J. Hawkins et al.
, n'I:lr+ n
PCA pump
lRate and bolus set]
Pneumatic connection
Table 1. Accuracy of PCA pumps in all three delivery modes. Bolus + Bolus Infusion infusion (YOof nominal volume delivered)
Long needle
electrical
connection
Measuring cylinder
ram
Graseby (average) Provider (average) IVAC
Interface
IRS232-
C
Comaliter
Fig. 1. A diagrammatic representation of the test system.
balance (RS232-C) routed through the interface box. The system is illustrated in Figure I . Using this system an assessment was made of the acuracy of 12 Graseby PCAS devices (Graseby Medical, Watford, UK), some of which had been in service in our hospital for the past 5 years. Three new Provider 5500 pumps (Abbott Ambulatory Infusion Systems, San Diego, USA) and one new IVAC PCAinfuser (IVAC Corporation, San Diego, USA) were also tested Each PCA device was set up to deliver water in three different delivery modes: (1) bolus only: 1 ml bolus doses according to requests from the computer; (2) infusion only: I m1.h-' infusion, no bolus requests; (3) bolus + infusion: 1 m1.h-I infusion plus 1 ml bolus doses according to requests from the computer. Each PCA device was studied for 24 h in each delivery mode. To produce results more representative of the use of PCA devices in the wards, the timing of bolus requests was based on clinical observations of patients using PCA for postoperative pain relief. The chosen profile is illustrated in Figure 2. The weight of infusate (water) delivered was converted to a volume by dividing by the density of the water a t the temperature in the laboratory (0.997 g.mI-I). Data were analysed using the Supercalc 4 spreadsheet programme where the accuracy of hourly delivery in the three modes was expressed as a percentage of the expected volume delivered. Statistical analysis was carried out on the data from the Graseby PCAS devices. Differences in accuracy among the three modes of delivery were compared using repeated measures analysis of variance. The data from the other devices were insufficient to warrant statistical analysis. To determine whether results may have been affected by evaporation of infusate, balance drift, or other equipment artifacts, the system was operated for 10 24 h periods during which no fluid was delivered.
Results When the system was run with no infusate being delivered the mean fluctuation was 0.OOg (SD = 0.08, range -0.09 g to + 0.14 g) over a 24 h period. In the worst possible case of
0-1 h 1-24 h 0-1 h 1-24 h 0-1 h 1-24 h
88.3 98.9 (1.72) 107.5 101.5 (1.18) 98.8 95.9 (9.83)*
51.9 97.6 (1.75) 102.6 99.4 (1.52) 81.2 97.3 (1.71)
88.7 98.3 (0.88) 99.3 99.3 (1.38) 96.1 97.8 (6.87)'
The values for the 12 Graseby PCAS and three Provider 5500 pumps are averaged. The values shown for the 1-24 h periods are expressed as mean (standard deviation). *These large standard deviations are explained in the discussion.
the maximum fluctuation occurring in the infusion-only mode (smallest amount delivered) this would only lead to a discrepancy of 0.58%. These fluctuations were therefore considered insignificant. With all the devices tested there was little difference in the hourly accuracy except in the first hour (Table I). The Graseby PCAS and IVAC PCAinfuser both underdelivered and the Abbott Provider 5500 overdelivered during this time. There was also a transient period of underdelivery with the Graseby and IVAC pumps following the 7 h period of no bolus requests, but delivery did not fall below 90% of that which was expected (see Fig. 3). The Abbott Provider 5500 delivered relatively accurately over the remaining 23 h period. If the two periods of inaccurate delivery described are excluded, then the average delivery was generally within 3% of expected in all modes. The exceptions were two of the Graseby PCAS devices which gave mean delivery rates of 95.7% and 94.2% in the infusion-only mode. Statistical analysis of the results for the Graseby PCAS device showed a difference among the modes in both the first hour and in the 1-24 h periods, the infusion-only mode giving the greatest underdelivery in both cases.
Discussion A fully automated system has been developed which can be used to test PCA devices. The system has been used to test
i\a
-k 4 g
3
Time ( h )
Fig. 2. Demand rate profile used to test PCA devices. The patient has a high request rate initially which declines to zero at night. An early morning peak is usually seen which settles to a low request rate.
Testing PCA devices
r loo
40
2o
t
t t I
0
100
l Graseby P C A S (mean)
I
I
I
I
I
I
4
8
12
16
20
24
4
-
8
I2
16
20
24
I00
7
IVAC PCAinfuser
20
I
0
use of the devices. It was found that they first primed the syringes by hand prior to placing them in the pump and we decided to use this same method in the knowledge that priming the system using the pump itself may have given a more accurate delivery in the first hour. However, if the syringe later required changing on the ward the latter approach would not be possible without either disconnecting the infusion or giving the patient a bolus dose. The reason for underdelivery in the hour following the 7 h period of no bolus requests is unclear. One explanation could be the development of increased compliance in the delivery system which has to be overcome prior t o accurate delivery rates being achieved. The significantly different accuracy of the Graseby PCAS in the three modes is likely to be due in part to the differing nominal values of infusate delivered. In the first hour only I ml was delivered in the infusion-only mode compared to 6ml and 7 m l in the bolus-only and combined modes respectively. This would lead to less inaccuracy in the latter two delivery modes, as was observed (Table l), as there would be greater tendency for the compliance changes and priming differences to be overcome within that hour. In the 1-24 h period, although the difference in the delivery between the modes was statistically significant, the actual percentage differences were small and clinically unimportant. The accuracy of the Abbott Provider 5500 is probably due to the fact that it operates as a peristaltic pump rather than a syringe driver. The system must be primed using the pump mechanism and there seems to be no effect of increasing compliance. On testing the W A C PCAinfuser it was found that in the second hour there was a failure to give one of the requested bolus doses, hence the large standard deviation value in the 1-24 h period. The software in this device limits bolus requests to a maximum of 10 in any 2 h period. The reason behind this may be that if any more than 10 requests are made then the manufacturers believe that there should be a review of the prescription. The number of bolus doses requested and the number given over the previous 24 h are displayed thereby giving an indication of a prescription problem. However, in the opinion of the authors, this arbitrary limit should not be included in the software. When a lockout period is chosen both prescriber and patient should be able to expect a bolus dose to be available if that period is exceeded. If a large number of requests are made then dosages should be reviewed but not limited by the PCA device. The visual display of requests on the device is a useful option. The system described has been shown to be useful in the testing of the accuracy of PCA devices and it has revealed some idiosyncrasies of PCA pump design. It will be possible to test all PCA devices available through appropriate modifications to the interface where required. This system is reliable, fully automated and could well be used in the future to test all PCA devices in a clinically relevant manner.
F8 Provider 5500 (mean)
0
695
I
I
I
I
I
I
4
8
12
16
20
24
Time ( h )
Fig. 3. The graphs show the actual volume delivered/expected volume delivered as a percentage for each hour. Only bolus ( 0 )and infusion (0)modes are shown for clarity.
a number of devices which were shown to be reasonably accurate i.e. within 10% of the bolus or continuous infusion dose selected [7]. In particular they were shown to be accurate in a clinically relevant delivery mode. However, unprimed syringe driver systems could underdeliver by more than 10% in the first hour. It has been suggested that bolus volumes should be accurate to within 10% of the selected dose. With continuous infusion modes the volumes should be accurate to within 5% between flow settings of 5-50 m1.h-I and 10% for all other settings [7]. It is surely more important that accuracy is greatest at flow rates used clinically (e.g. 1 m1.h-I in .our centre) especially as more concentrated solutions are used and thus discrepancies can have greater clinical effects. The low volumes delivered in the first hour by the Graseby pumps, and to a lesser extent the W A C pump, can be explained if we consider the method used to set up the PCA devices. A study was made of the method used by the nurses in our hospital in an effort to reflect further clinical
References [ I ] SECHZER PH.Objective measurement of pain. Anesthesiology 1968; 2 9 209-10. [2]
READYLB. Patient-controlled analgesia - does it provide more than comfort? Canadian Journal of Anaesthesia 1990; 37: 719-21.
696
D.J. Hawkins et al.
[3] RAPP RP, BIVINS BA, LI'ITRELL RA, FOSTER TS. Patient-controlled analgesia: a review of effectiveness of therapy and an evaluation of currently available devices. DICP. 1989; 2 3 899-904. [4] OWEN H, GLAVIN RJ, REEKIE RM, TREW AS. Patient-controlled analgesia. Experience of two new machines. Anaesthesia 1986; 41: 1230-5. [S] BALDWINAM, ILSLEY AH, KLUGERMT, OWENH. Assessment
of a new infusion pump for epidural PCA. Anaesthesia and Intensive Care 1991; 1 9 246-50. [6] JACKSON IJB, SEMPLE P, STEVENS JD. Evaluation of the Graseby PCAS. A clinical and laboratory study. Anaesthesia 1991; 46: 482-5. [7] Patient-controlled analgesia infusion pumps. Health Devices 1988; 17: 137-67.
An automated system for testing the accuracy of patient-controlled analgesia devices
D. J . HAWKINS, D. R . D. ROBERTS, A . H. ILSLEY, J. L. PLUMMER
AND
H . OWEN
Summary A system was developed to test the accuracy of patient-controlled analgesia devices in situations simulating clinical use. Bolus requests are made automatically at predetermined intervals, and the infusate delivered is measured and recorded without the need for operator presence. To ensure clinical relevance, the bolus request times used in this study corresponded to a pattern typical of those requested by patients on the ward. Graseby, Abbott Provider 5.500 and W A C patient-controlled analgesia devices were tested and found to deliver reasonably accurately over a 24 h period. However, when an infusion was started in an unprimed system or a f e r a period of no bolus requests in a bolus-only mode the Graseby and W A C machines under-delivered. This system provides a means of testing patient-controlled analgesia devices operating in any delivery mode.
Key words Analgesia; patient-controlled. Equipment; infusion pumps, computer.
Over the last 20 years, patient-controlled analgesia (PCA) has developed from an investigative tool [I] to a well-used and desirable form of postoperative pain relief [2]. The choice of suitable PCA devices has increased accordingly although . . . ‘many are available for use without a single publication to document the safety and utility of the device in the routine patient care situation’ [3]. Where articles have been published reviewing these devices, most report their accuracy [3-61 assessed volumetrically, or more usually gravimetrically, from collection of infusate delivered. In these studies the output was measured manually, and continuous infusion and bolus doses were tested separately with no effort being made to simulate clinical use. Testing a large number of PCA devices under simulated clinical conditions would be very time consuming and to request bolus doses at timed intervals, with and without a background infusion, would be even more laborious. One solution is a computer-controlled system where bolus requests can be made by the computer and the weight of infusate delivered recorded automatically. The aim of this study was to develop a system for the automated assessment of PCA devices with particular attention to their accuracy in a simulation of clinical use. Three types of electronic PCA devices were studied using this system.
Methods A TIOOO lap-top computer (Toshiba corporation, Tokyo, Japan) was used to control a system which could initiate demands and gravimetrically measure the fluid delivered by the PCA devices. Narrow bore plastic tubing connected the PCA device to a long 14G needle which rested on the bottom of a measuring cylinder below a layer of light oil introduced to reduce inaccuracies due to evaporation. Fluid delivered by each PCA device was continuously weighed on a calibrated Mettler PM2000 electronic balance (Mettler Instruments, AG Zurich, Switzerland). The computer was programmed to interrogate the balance hourly; the weight and time were then automatically recorded on computer disk. The computer program simulated the bolus requests of a patient via a purpose-built interface box connected to the PCA device. When a bolus was requested the interface box activated the PCA device either directly by an electrical connection or by a pneumatic connection (for the Graseby PCAS). The pneumatic pressure was provided by a Maxima pneumatic aquarium pump (Ralph C. Hagen, Leeds, UK). Communication was via serial connections made between the computer and interface box and between the computer and the electronic
D.J. Hawkins, FRCAnaes, D.R.D. Roberts, FRCAnaes, Senior Registrars, A.H. Ilsley, PhD, J.L. Plummer, PhD, Principal Hospital Scientists, H. Owen, MD, FANZCA, Senior Staff Specialist, Department of Anaesthesia and Intensive Care, Flinders Medical Centre, Bedford Park, South Australia 5042, Australia. Correspondence should be addressed to Dr H. Owen. Accepted 23 December 1991. 0003-2409/92/080693 + 04 $03.00/0
@ 1992 The Association of Anaesthetists of Gt Britain and Ireland
693
694
D.J. Hawkins et al.
, n'I:lr+ n
PCA pump
lRate and bolus set]
Pneumatic connection
Table 1. Accuracy of PCA pumps in all three delivery modes. Bolus + Bolus Infusion infusion (YOof nominal volume delivered)
Long needle
electrical
connection
Measuring cylinder
ram
Graseby (average) Provider (average) IVAC
Interface
IRS232-
C
Comaliter
Fig. 1. A diagrammatic representation of the test system.
balance (RS232-C) routed through the interface box. The system is illustrated in Figure I . Using this system an assessment was made of the acuracy of 12 Graseby PCAS devices (Graseby Medical, Watford, UK), some of which had been in service in our hospital for the past 5 years. Three new Provider 5500 pumps (Abbott Ambulatory Infusion Systems, San Diego, USA) and one new IVAC PCAinfuser (IVAC Corporation, San Diego, USA) were also tested Each PCA device was set up to deliver water in three different delivery modes: (1) bolus only: 1 ml bolus doses according to requests from the computer; (2) infusion only: I m1.h-' infusion, no bolus requests; (3) bolus + infusion: 1 m1.h-I infusion plus 1 ml bolus doses according to requests from the computer. Each PCA device was studied for 24 h in each delivery mode. To produce results more representative of the use of PCA devices in the wards, the timing of bolus requests was based on clinical observations of patients using PCA for postoperative pain relief. The chosen profile is illustrated in Figure 2. The weight of infusate (water) delivered was converted to a volume by dividing by the density of the water a t the temperature in the laboratory (0.997 g.mI-I). Data were analysed using the Supercalc 4 spreadsheet programme where the accuracy of hourly delivery in the three modes was expressed as a percentage of the expected volume delivered. Statistical analysis was carried out on the data from the Graseby PCAS devices. Differences in accuracy among the three modes of delivery were compared using repeated measures analysis of variance. The data from the other devices were insufficient to warrant statistical analysis. To determine whether results may have been affected by evaporation of infusate, balance drift, or other equipment artifacts, the system was operated for 10 24 h periods during which no fluid was delivered.
Results When the system was run with no infusate being delivered the mean fluctuation was 0.OOg (SD = 0.08, range -0.09 g to + 0.14 g) over a 24 h period. In the worst possible case of
0-1 h 1-24 h 0-1 h 1-24 h 0-1 h 1-24 h
88.3 98.9 (1.72) 107.5 101.5 (1.18) 98.8 95.9 (9.83)*
51.9 97.6 (1.75) 102.6 99.4 (1.52) 81.2 97.3 (1.71)
88.7 98.3 (0.88) 99.3 99.3 (1.38) 96.1 97.8 (6.87)'
The values for the 12 Graseby PCAS and three Provider 5500 pumps are averaged. The values shown for the 1-24 h periods are expressed as mean (standard deviation). *These large standard deviations are explained in the discussion.
the maximum fluctuation occurring in the infusion-only mode (smallest amount delivered) this would only lead to a discrepancy of 0.58%. These fluctuations were therefore considered insignificant. With all the devices tested there was little difference in the hourly accuracy except in the first hour (Table I). The Graseby PCAS and IVAC PCAinfuser both underdelivered and the Abbott Provider 5500 overdelivered during this time. There was also a transient period of underdelivery with the Graseby and IVAC pumps following the 7 h period of no bolus requests, but delivery did not fall below 90% of that which was expected (see Fig. 3). The Abbott Provider 5500 delivered relatively accurately over the remaining 23 h period. If the two periods of inaccurate delivery described are excluded, then the average delivery was generally within 3% of expected in all modes. The exceptions were two of the Graseby PCAS devices which gave mean delivery rates of 95.7% and 94.2% in the infusion-only mode. Statistical analysis of the results for the Graseby PCAS device showed a difference among the modes in both the first hour and in the 1-24 h periods, the infusion-only mode giving the greatest underdelivery in both cases.
Discussion A fully automated system has been developed which can be used to test PCA devices. The system has been used to test
i\a
-k 4 g
3
Time ( h )
Fig. 2. Demand rate profile used to test PCA devices. The patient has a high request rate initially which declines to zero at night. An early morning peak is usually seen which settles to a low request rate.
Testing PCA devices
r loo
40
2o
t
t t I
0
100
l Graseby P C A S (mean)
I
I
I
I
I
I
4
8
12
16
20
24
4
-
8
I2
16
20
24
I00
7
IVAC PCAinfuser
20
I
0
use of the devices. It was found that they first primed the syringes by hand prior to placing them in the pump and we decided to use this same method in the knowledge that priming the system using the pump itself may have given a more accurate delivery in the first hour. However, if the syringe later required changing on the ward the latter approach would not be possible without either disconnecting the infusion or giving the patient a bolus dose. The reason for underdelivery in the hour following the 7 h period of no bolus requests is unclear. One explanation could be the development of increased compliance in the delivery system which has to be overcome prior t o accurate delivery rates being achieved. The significantly different accuracy of the Graseby PCAS in the three modes is likely to be due in part to the differing nominal values of infusate delivered. In the first hour only I ml was delivered in the infusion-only mode compared to 6ml and 7 m l in the bolus-only and combined modes respectively. This would lead to less inaccuracy in the latter two delivery modes, as was observed (Table l), as there would be greater tendency for the compliance changes and priming differences to be overcome within that hour. In the 1-24 h period, although the difference in the delivery between the modes was statistically significant, the actual percentage differences were small and clinically unimportant. The accuracy of the Abbott Provider 5500 is probably due to the fact that it operates as a peristaltic pump rather than a syringe driver. The system must be primed using the pump mechanism and there seems to be no effect of increasing compliance. On testing the W A C PCAinfuser it was found that in the second hour there was a failure to give one of the requested bolus doses, hence the large standard deviation value in the 1-24 h period. The software in this device limits bolus requests to a maximum of 10 in any 2 h period. The reason behind this may be that if any more than 10 requests are made then the manufacturers believe that there should be a review of the prescription. The number of bolus doses requested and the number given over the previous 24 h are displayed thereby giving an indication of a prescription problem. However, in the opinion of the authors, this arbitrary limit should not be included in the software. When a lockout period is chosen both prescriber and patient should be able to expect a bolus dose to be available if that period is exceeded. If a large number of requests are made then dosages should be reviewed but not limited by the PCA device. The visual display of requests on the device is a useful option. The system described has been shown to be useful in the testing of the accuracy of PCA devices and it has revealed some idiosyncrasies of PCA pump design. It will be possible to test all PCA devices available through appropriate modifications to the interface where required. This system is reliable, fully automated and could well be used in the future to test all PCA devices in a clinically relevant manner.
F8 Provider 5500 (mean)
0
695
I
I
I
I
I
I
4
8
12
16
20
24
Time ( h )
Fig. 3. The graphs show the actual volume delivered/expected volume delivered as a percentage for each hour. Only bolus ( 0 )and infusion (0)modes are shown for clarity.
a number of devices which were shown to be reasonably accurate i.e. within 10% of the bolus or continuous infusion dose selected [7]. In particular they were shown to be accurate in a clinically relevant delivery mode. However, unprimed syringe driver systems could underdeliver by more than 10% in the first hour. It has been suggested that bolus volumes should be accurate to within 10% of the selected dose. With continuous infusion modes the volumes should be accurate to within 5% between flow settings of 5-50 m1.h-I and 10% for all other settings [7]. It is surely more important that accuracy is greatest at flow rates used clinically (e.g. 1 m1.h-I in .our centre) especially as more concentrated solutions are used and thus discrepancies can have greater clinical effects. The low volumes delivered in the first hour by the Graseby pumps, and to a lesser extent the W A C pump, can be explained if we consider the method used to set up the PCA devices. A study was made of the method used by the nurses in our hospital in an effort to reflect further clinical
References [ I ] SECHZER PH.Objective measurement of pain. Anesthesiology 1968; 2 9 209-10. [2]
READYLB. Patient-controlled analgesia - does it provide more than comfort? Canadian Journal of Anaesthesia 1990; 37: 719-21.
696
D.J. Hawkins et al.
[3] RAPP RP, BIVINS BA, LI'ITRELL RA, FOSTER TS. Patient-controlled analgesia: a review of effectiveness of therapy and an evaluation of currently available devices. DICP. 1989; 2 3 899-904. [4] OWEN H, GLAVIN RJ, REEKIE RM, TREW AS. Patient-controlled analgesia. Experience of two new machines. Anaesthesia 1986; 41: 1230-5. [S] BALDWINAM, ILSLEY AH, KLUGERMT, OWENH. Assessment
of a new infusion pump for epidural PCA. Anaesthesia and Intensive Care 1991; 1 9 246-50. [6] JACKSON IJB, SEMPLE P, STEVENS JD. Evaluation of the Graseby PCAS. A clinical and laboratory study. Anaesthesia 1991; 46: 482-5. [7] Patient-controlled analgesia infusion pumps. Health Devices 1988; 17: 137-67.
