This site is intended for health professionals only!
Department of Pharmacy, Section of Pharmacotherapy and Pharmaceutical Care,
University of Groningen, The Netherlands
Bryony Dean Franklin
Director, Centre for Medication Safety and Service Quality, Imperial Healthcare NHS Trust and School of Pharmacy, University of London, UK
Medication errors are unacceptably common. They can occur during each step in the process of providing medication to hospitalised patients: prescribing, dispensing, preparation, administration and monitoring. Intravenous medications are essential in the therapeutic management of many patients, especially those who are critically ill.
However, providing intravenous drug therapy is a complex process. Much higher error rates have been reported than for non-intravenous doses, largely due to the additional complexity involved in the process of preparation and administration of intravenous doses.
For studies using the gold standard method of observation of medication administration, published error rates vary from 18% to 173% of doses given.1 A recent systematic review estimated the probability of making at least one error in the preparation and administration of intravenous medication to be 0.73, with the highest percentages of errors occurring at the reconstitution and administration steps.2
To reduce these errors, the introduction of automated infusion devices or ‘smart pumps’ has been advocated.3 These incorporate software that checks programmed doses against preset limits for each drug and clinical location, using customisable ‘drug libraries’. Limits may be ‘soft’ (in which case they can be over-ridden) or hard (in which case they cannot).
Pumps may include a range of other features such as being networked – either hard-wired or wireless – and integrated with other electronic systems including electronic prescribing, bar code-assisted medication administration, and medication administration records. Most smart pumps also allow administration data, such as number and types of overrides, to be downloaded for quality control purposes.
This paper presents the evidence for the effectiveness of such technology and discusses the implications for European hospital pharmacists.
Evidence of smart infusion pumps’ effectiveness
A 2005 study in one US hospital found that, of 426 observed infusions using conventional pumps, 67% (285) were associated with one or more errors.4 There were 37 rate errors, but only one of these was judged to be potentially preventable using the smart infusion pumps that were available at the time of the study. More were judged to be potentially preventable with more advanced pumps with functionality to integrate with other electronic systems in the hospital, such as those now available.
A US study in a paediatric intensive care unit found discrepancies between what was prescribed using computerised physician order entry (CPOE) and what was administered using smart pump technology in 24% out of 296 observed drug infusions and 42% of 231 fluid infusions.5 The majority of these discrepancies were unauthorised orders or omissions; few were dosing errors. However, it was not possible to ascertain which of the CPOE records and the infusions administered were actually ‘correct’, as it was common practice in the hospital concerned for prescribers to give verbal orders without details being documented on the CPOE system. These findings are therefore difficult to interpret.
A number of papers report analyses of data downloaded from smart infusion pumps about reprogramming events in response to alerts. A study of nearly two years in a US critical care unit identified 970 alerts, of which 14% were judged to have potentially prevented patient harm.6
Reprogramming events and overrides in the administration of high-risk medications to both adult and paediatric ICU patients were investigated in another US study.7 This suggested that smart pumps were particularly beneficial for paediatric patients. Another study concluded that using smart pumps potentially prevented both underdosing and overdosing errors in anticoagulants.8 However, descriptive non-experimental studies of this type do not tell us what would have happened in the absence of the smart pump.
We identified four studies investigating the effectiveness of smart infusion pumps using an experimental design (see Table 1). Three were carried out in clinical practice and one in a simulated environment. Based on these studies, evidence for smart pumps’ impact on patient safety is so far inconclusive.
A study published in 2005 showed that the use of smart pump decision support had no effect on the prevalence of serious medication errors, although the authors note that the decision support software was used for only 75% of infusions in the intervention phase.9
A second, published the same year, demonstrated a significant reduction in medication errors following the introduction of a combination of smart pumps, standard infusion concentrations and redesigned medication labels.10
A study in 2007 suggested that the use of smart pumps had no effect on the rate of preventable adverse drug events associated with intravenous medication.11 However, the three studies are difficult to compare, as they were carried out in very different clinical areas, with different types of infusion pumps.
Furthermore, each study used different outcome measures and data collection methods. One used a comprehensive approach to identifying adverse events using medical record review, incident reports and log data from the pumps,9 another used a retrospective review of medical notes,11 and the third based its data on incident reports.10 Incident report data is known to be subject to considerable under-reporting.12
A simulation study showed that smart pumps prevented wrong dose errors when using hard limits.13 However, there was no significant effect when using soft limits, as nurses overrode the alerts. Pumps with barcode technology prevented administration to the wrong patient. It was also found that the choice of pump did not make any difference in the prevention of incorrect drug errors, possibly because of a lack of integration of the pump with other medication management technologies.
Although the results of this simulation study may not be entirely applicable in general if people act differently in ‘real’ clinical practice, this study showed the potential of smart pumps to improve medication safety, but that this depends on the pump specifications and how the software is set up.
In summary, no studies using experimental designs reported an increase in medication error rates, but results are inconsistent about whether or not medication error rates are reduced. This is likely to depend on how the pumps are implemented and used in clinical practice.
Furthermore, all studies to date have been carried out in North America. Their results therefore may or may not be applicable to Europe and other parts of the world due to differences in clinical practice. For example, in the USA, most hospital pharmacies prepare all infusions and will include a label with information on the rate at which to set the smart pump, whereas this is not generally the case in European hospitals.
Introducing and using smart pumps in the clinical setting
As with most similar technologies, various issues must be taken into account when introducing smart pumps in order to fully realise their benefits. An important prerequisite for the introduction of smart pumps is standardisation of the concentrations and volumes of drugs to be administered, to create a suitable drug library. Such standardisation is generally considered to be desirable even in the absence of smart pumps,14 but may require a considerable amount of work in gaining consensus if a wide range of practices exist.
Such a standardisation process is not a one-off event, but has to be ongoing to incorporate changes in practice and product availability. Any mismatches between the infusions prescribed, the products available and the information in the drug library will create a high risk of error. All changes must also be communicated to front-line staff.
A key implementation question is whether the whole organisation is converted to smart pumps at the same time (the ‘big bang’ approach), whether one clinical unit is converted at a time, or whether pumps are replaced piecemeal as replacements are required. The latter approach is likely to be the most risky, resulting in a range of pumps being used simultaneously, with potential confusion for users. Whatever approach is adopted, suitable administrative oversight and multidisciplinary leadership will be required.
The choice of hard versus soft overrides needs careful consideration. Hard overrides are potentially useful, but if used inappropriately will result in non-compliance with the safety software, with staff adopting various work-arounds and shortcuts to administer the medication required.15 Full clinical consultation is needed.
A related issue is whether or not to make the drug library mandatory. If the drug library is set to be mandatory, it is not possible to administer ‘one-off’ infusions that are not in the library – for example, for clinical trial medication or for specific clinical indications. However, if it is possible to bypass the drug library, staff may choose this option rather than using the safety features of the software, resulting in the full benefits of the technology being lost.9
Different drug libraries can be set for different clinical areas. Each ward can be defined as its own clinical area, or different patient populations can be defined, each with its own drug library. The best option is likely to depend on the epidemiology of infusion use and the amount of variation between and within ward areas.16
If reporting functionality is enabled, a range of data can be downloaded, including the drugs and strengths most commonly used, details of overrides used for soft limits, and the programming errors prevented. Using such data requires considerable resource and dedication, but has the potential to be enormously useful in supporting a continuous quality improvement approach to intravenous medication administration.16 Such data can be used to identify modifications that are required to drug libraries, standardised infusions and the hard and soft limits adopted, as well as to identify areas where further training and support are needed for frontline staff.
Smart pumps offer potentially safer infusion of intravenous medication. However, to achieve their true potential, pumps and associated drug libraries need to be introduced with due consideration of a range of issues, with ongoing monitoring of how they are actually being used in practice.
The future is likely to bring further development of pumps, especially improving the easy integration of smart pumps into existing electronic applications in hospitals resulting in truly closed loop systems. Ongoing evaluation is needed to ensure that benefits are realised.
1. Franklin BD, Taxis T and Barber N. [letter] BMJ 2009;338:b1814.
2. McDowell SE et al. Quality and Safety in Health Care 2010;19:341-345.
3. Forni A et al. Current Drug Safety 2010;5:13-18.
4. Husch M et al. Quality and Safety in Health Care 2005;14:80-86.
5. Russell RA et al. Quality and Safety in Health Care 2010;19(Suppl.3):i31-i35.
6. Evans RS et al. Stud Health Technol Inform. 2010;160(Pt 1):734-8.
7. Hennings S et al. Hospital Pharmacy 2010; 45:464-471.
8. Fanikos J et al. American Journal of Cardiology 2007; 99(7): 1002-5.
9. Rothschild JM et al. Critical Care Medicine 2005;33:533-540.
10. Larsen GY et al. Pediatrics 2005; 116:e21-e25.
11. Nuckols TK et al. J Gen Inter Medicine 2007;23(Suppl. 1):41-45.
12. Flynn EA et al. Am J Health Syst Pharm. 2002 Mar 1;59(5):436-46.
13. Trbovich PL et al. Quality and Safety in Health Care 2010;19:430-434.
14. Keeling P et al. British Journal of Nursing (Intravenous Supplement) 2010;30:S30-S33.
15. McAlearney AS et al. Journal of Patient Safety 2007;3:75-81.
16. Breland BD. Am J Health-System Pharmacy 2010;67:1446-55.