Integrated monitoring and analysis for early warning of patient deterioration
Editor—We read with interest the results from the review by Tarassenko and colleagues1 describing the real-time automated system, BioSign and its role as part of a strategy to detect at-risk patients to trigger timely interventions by the Medical Emergency Team.It is clear that with the lack of level 2/3 beds available there needs to be a system in place to allow Critical Care outreach services to function optimally. This will allow the moderately sick patients to be managed on the ward and allow the sickest patients to be admitted to the scarce ICU beds. There have been numerous studies which have shown that abnormalities in basic physiological observations are present in patients before they are admitted to ICU2 or have a cardiac arrest.3 The Early Warning Scores (EWS) have evidence4 to support their use; however, if these ‘track and trigger’ systems are to be effectively used the stimulus to pull the trigger needs to be timely and correctly interpreted. The BioSign system may allow the introduction of the EWS onto wards and solve some of the hurdles and objections to relying on these scoring systems.
In the current staffing of some surgical wards it appears that the recording of these basic physiological parameters is often delegated to inexperienced healthcare assistants or the interval between observations can be significant. The accurate recording of these vital observations is crucial to the value of EWS, otherwise these scores are meaningless and no action is taken.
We looked at the introduction of the Modified Early Warning Score (MEWS) on our surgical wards at Peterborough General Hospital as part of the Critical Care Outreach Programme.5 This Outreach service is provided by a Patient Outreach Team (PORT) during the hours of 08:00–20:00, 7 days a week. These critical care trained nursing staff review at risk patients who are referred by the ward nursing or medical staff. The PORT clinically manages the patient and become the ‘trigger’ for the involvement of the Intensive Care Team if required. The ‘tracking’ system is the MEWS and when the score is recorded as four or more the patient is to be referred to either medical staff or the PORT or both. The PORT introduced new ward observation charts to allow a systematic recording of the MEWS and prompts to remind staff which scores required further action.
After the introduction of these scoring systems onto the wards we performed an audit in November 2005 to try to demonstrate whether the recording of the MEWS on postoperative patients was being performed. Our audit looked at 30 patients on General, Vascular and Orthopaedic wards at Peterborough District Hospital. The observation charts were reviewed whilst these patients were on the wards and it was noted whether the MEWS was recorded and if recorded whether it was accurately calculated by ward staff. The patients were all between 1 and 10 days after operation, ASA grades (II–IV), and were on the ward at the times the data were collected.
The results revealed that MEWS was documented in 69% of patients and in those patients only 42% had a correct MEWS. This implies that an accurate MEWS was recorded in only 29% of postoperative surgical patients at the time the audit was performed.
If all medical and nursing staff are not educated on the value and purpose of these physiological parameters and the EWS used, then these ‘track and trigger’ systems may well ‘misfire’ and be a meaningless tool. The BioSign's automated function allows it to automatically reproduce EWS and alert nursing and medical staff to an early deterioration. This would correct any human variations in both the accuracy of the score and the frequency of clinical observation recording. It is also worth commenting that there can be no substitute for bedside clinical assessment and these systems may also prompt earlier such assessments in sick patients.
F. Ismail
M. Davies*
Peterborough, UK
*E-mail: mtdavies{at}lineone.net
Editor—Tarassenko and colleagues1 comment on the apparent absence of any attempts to automate the process of calculating early warning scores (EWS) for identifying sick patients. For approximately 2 yr, we have been developing a system to do just this in a general ward environment. This has now been successfully implemented in our medical and surgical assessment units, and its extension to the rest of the hospital is being planned. Within our medical assessment unit alone, we are now collecting approximately 10 000 complete, vital signs datasets per month.
The personal digital assistant (PDA) based system collects all commonly recorded vital signs data and automatically calculates the EWS6 and thus removes the need for healthcare staff to know the weightings for individual components of the score. Unlike the BioSign system, nurses using our system are also able to input data on the patient's mental status and urine output. The system utilizes a wireless local area network (W-LAN) to integrate the raw physiology data, the EWS, vital signs charts and oxygen therapy records with the hospital's laboratory and activity databases. These data are then instantaneously available to any member of the hospital healthcare team with access to the W-LAN or intranet. Additional important benefits are that the vital signs charts are legible, up-to-date, accurate and date- and time-stamped.
As part of the development of the system, we undertook a comparative study to investigate the accuracy of, and time taken to calculate, EWS for both the PDA system and the traditional pen and paper method.7 When the PDA was used, there was improved accuracy of EWS calculation, and a reduction in the time required to chart and calculate an EWS, compared with the traditional pen and paper method. An additional finding (unpublished data) was that the time required for EWS calculation using pen and paper increased with the severity of illness.
The mean time (in seconds) taken to calculate and chart an EWS was shorter using PDA (tVP) than pen and paper (tPP) for five patient scenarios (Fig. 1). Scenarios 1 and 2 were of equal severity of illness (i.e. identical EWS), whereas the EWS scores for scenarios 3, 4 and 5 increased progressively. All participants produced EWS scores for the five scenarios using both methods in randomized order and compared using repeated measures analysis of variance on the time to complete each scenario. The method, order in which it was used first and the scenario (1–5) were each treated as fully crossed fixed effects in a factorial model, with participants treated as random effects nested within order of method. Variability between participants was highly significant (P<0.001) and accounted for 52% of the variability in response times. The methods (P<0.001) and the scenario (P<0.001) were significant factors in determining times. A significant interaction between these two factors indicated that the difference in times for each scenario depended upon the method (P<0.001). There was no evidence that the order in which the methods were used (paper first or PDA first) had any main effect upon times (P=0.757). There was no evidence that this factor interacted with method (P=0.339), scenario (P=0.281) or the interaction between these two factors (P=0.414).
|
The plot of mean times for the two methods showed evidence of an equivalent learning effect for both input methods (a similar reduction in mean time between scenario 1 and scenario 2 was observed for both methods). Thereafter, the mean time taken using the PDA remained fairly constant, whereas pen and paper took longer for the more complex scenarios.
Declaration of interest
The PDA system referred to, VitalPACTM, is a collaborative development of The Learning Clinic Ltd and Portsmouth Hospitals NHS Trust.
G. Smith*
D. Prytherch
H. Peet
P. Featherstone
P. Schmidt
D. Knight
K. Stewart
B. Higgins
Portsmouth, UK
*E-mail: Gary.smith{at}porthosp.nhs.uk
Editor—We are very grateful to Drs Ismail and Davies as well as to Prof. Smith and colleagues for their comments on our review1 describing the real-time automated alerting system, BioSign and its role as part of a strategy to detect at-risk patients. When we wrote our review, there were no published reports on the system developed by Smith and colleagues to collect vital signs data at the bedside using PDAs.
In a paper published6 since our review, Smith and colleagues argue that ‘the ability of "track and trigger" systems to influence the incidence of adverse outcomes is governed by the nature and frequency of vital signs observations’ (our italics). They cite this as one reason why the MERIT trial8 failed to demonstrate the value of Medical Emergency Teams: ‘infrequent recording of vital signs data may partly explain why the outcomes in the control and intervention arms were similar’.6 Drs Ismail and Davies show that this is indeed the Achilles heel of score-based alerting systems, with only 29% of patients in their study having correct, timely values charted. The PDA-based system described by Smith and colleagues may be 15–30 s per patient faster than paper recording but is still subject to human error and vulnerable if staff fail to make measurements.
Even patients who are monitored continuously may be observed infrequently and unobserved monitors cannot convey any patient benefit.9 With current nursing procedures on acute wards, significant changes in vital signs are missed; for example, worst-case analysis of the ACADEMIA study3 shows that, even with 4-hourly observations, as many as 94% of the occurrences of significant clinical deterioration before an adverse event could be missed: only 12 of the 209 patients (5.7%) with antecedents in their vital signs and without a ‘do not attempt resuscitation’ (DNAR) order had antecedents for more than 4 h before the adverse event [cardiac arrest, death or intensive care unit (ICU) admission].
Thus, conventional monitoring and alerting alone cannot be used as a reliable means of identifying clinical deterioration. Alarms on single-channel monitors are simply not robust enough and are frequently switched off. Tsien and Fackler10 report a Positive Predictive Value (percentage of true alerts with respect to the total number of alerts) as low as 3% in ICUs. There are two possible solutions to this problem: either increase the frequency of observations or improve the robustness of the alerting. The former is not economically viable outside of the ICU; we have concentrated on the second option instead and developed BioSign as a means of delivering such an improvement. Through a combination of careful signal processing and data fusion across five vital signs, we were able to increase the Positive Predictive Value of BioSign alerts to 95% (652 out of 690 transitions from ‘normal’ to ‘abnormal’).9
There is one other fundamental difference between score-based alerting systems and BioSign. All published early warning scoring systems have been generated using expert opinion, and the cut-offs between the different scores for each variable do not reflect accurate risk stratification. BioSign is a data-driven technique, which estimates the probability that the currently monitored patient's data are the same as the data recorded from a group of typical high-risk patients. A BioSign index of 3.0 indicates that there is a probability of 0.05 that the vital signs of the patient being monitored are drawn from the same distribution as those recorded from the typical high-risk patients. As the Biosign index represents a probability, the percentage of time a monitored group of patients will trigger an alert can be estimated. As a result, the expected call-out rate for the Emergency or Outreach Team can be determined in advance.
L. Tarassenko*
A. Hann
D. Young
Oxford, UK
*E-mail: lionel{at}robots.ox.ac.uk
References
1 Tarassenko L, Hann A, Young D. Integrated monitoring and analysis for early warning of patient deterioration. Br J Anaesth 2006; 97:64–8
2 Cullinane M, Findlay G, Hargraves C, Lucas S. An Acute Problem2005.London National Confidential Enquiry into Patient Outcome and Death
3 Kause J, Smith G, Prytherch D, et al. A comparison of antecedents to cardiac arrest, deaths and emergency intensive care admissions in Australia and New Zealand, and the United Kingdom—the ACADEMIA study. Resuscitation 2004; 62:275–82[CrossRef][Web of Science][Medline]
4 Goldhill DR, McNarry AF, Mandersloot G, McGinley A. A physiologically-based early warning score for ward patients: the association between score and outcome. Anaesthesia 2005; 60:547–53[CrossRef][Web of Science][Medline]
5 Department of Health. Comprehensive Critical Care: A Review of Adult Critical Care Services2000.London Department of Health
6 Smith GB, Prytherch DR, Schmidt P, et al. Hospital-wide physiological surveillance—a new approach to the early identification and management of the sick patient. Resuscitation 2006; 71:19–28[CrossRef][Web of Science][Medline]
7 Prytherch D, Smith GB, Schmidt P, et al. Calculating early warning scores—a classroom comparison of pen and paper and hand-held computer methods. Resuscitation 2006; 70:173–8[CrossRef][Web of Science][Medline]
8 Hillman K, Chen J, Cretikos M, et al. Introduction of the medical emergency team (MET) system: a cluster-randomised controlled trial. Lancet 2005; 365:2091–7[CrossRef][Web of Science][Medline]
9 Watkinson PJ, Barber VS, Price JD, et al. A randomised controlled trial of the effect of continuous electronic physiological monitoring on the adverse event rate in high risk medical and surgical patients. Anaesthesia 2006; 61:1031–9[CrossRef][Web of Science][Medline]
10 Tsien CL and Fackler JC. Poor prognosis for existing monitors in the intensive care unit. Crit Care Med 1997; 25:614–9[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||