Background
Medication errors (MEs) are common in pediatric inpatient populations, and the risk of potential adverse drug events is significant in neonates, particularly in neonatal intensive care units (NICUs) [
1‐
3]. Neonates are exposed to a higher risk of harm from MEs because of their rapidly changing body size and physical development, challenges to communicate with care providers, and more limited internal reserves to compensate for errors [
4‐
6]. Also, the medication-use process in NICU is particularly complex because of the wide use of intravenous (IV) administration routes, weight-based small dosages, multiple calculations and dilutions, common off-label use, and the use of unlicensed drugs [
2,
7‐
9]. MEs resulting in 10-fold, 100-fold, and even 1000-fold overdoses have been reported within the NICU settings, while such large deviations from the intended dose are less common in adult populations [
10‐
14]. Moreover, many intravenously administered high-alert medications, such as opioids, insulin, vasoactive drugs, and parenteral nutrition, are used in NICU settings [
2,
3,
7,
15]. As high-alert drugs bear a heightened risk for harm when used in error, proactive risk management strategies should be used to optimise these medication-use processes of neonatal patients [
7,
15,
16].
The manual adjustment of infusion rates for each patient has been identified as an especially high-risk task of the IV medication-use process [
7,
11,
17]. Smart infusion pumps with dose error reduction software (DERS) and associated drug libraries are designed to provide users with decision support in order to identify programming errors before starting the infusion [
18]. Drug specific dosing limits can be placed to prevent both overdosing (upper limits) and underdosing (lower limits). While soft limits are intended to advise the user of potential errors and can be overridden, hard limits force functions to ensure that the facility-established medication-specific parameters are not exceeded. In the literature, unnecessary alerts resulting from poorly chosen dosing limits have been reported to contribute to alert fatigue among healthcare professionals using smart infusion pumps [
18‐
25]. As a result, new medication safety risks arise from insufficient compliance in drug library use, and high override rates of soft limits have been identified. Other barriers to optimize the use of smart infusion pumps include limitations in pump capabilities, availability of pumps, programming workflow, associated risks with secondary infusions, pump data analysis, and deficiencies related to drug library use and updates (e.g., omitting certain drugs or IV fluids) [
18,
26,
27].
To maximise the benefits of smart infusion pumps as a systemic defence and risk-reduction strategy, the dosing limits should be carefully optimised for each drug, patient group and care area before implementation [
16,
18,
19,
21,
22,
28‐
31]. However, the scientific evidence about suitable methods for optimising the dosing limits in any patient care setting prior to their implementation in drug libraries is currently limited. The studies related to smart infusion systems mainly focus on the assessment of drug library compliance among caregivers, rate of triggered alerts, and soft limit override rates by using retrospective data collected from smart pump records [
19,
22]. Also, studies exploring smart pumps in NICU [
21] or wider hospital settings treating neonates [
23‐
25,
32,
33] have focused on describing the building of drug libraries in a general level, and retrospectively evaluating drug library compliance or triggered alerts. The principles for setting dosing limits prior to implementation have not been described in detail. Overall, the evidence related to implementing smart infusion systems in the NICU settings is limited [
21]. Therefore, the aim of this study was to develop a method for defining and assessing the optimal dosing limits in a NICU drug library. First, the utility of reported MEs in developing test patient cases to simulate potential programming errors was explored. Second, the alerts caused by the test cases were investigated to conclude the optimal dosing limits of the drug libraries.
Discussion
To the best of our knowledge, this is the first study aiming to optimise drug library dosing limits in smart infusion pumps prior to their implementation in a NICU environment. Our research was based on the systems approach to preventive medication safety risk management stating that risks should be identified and managed proactively before they reach the patient [
16]. The findings of our study support the use of hospitals’ own ME reports and the existing literature to identify risks associated with wrong infusion rate and to optimise drug library dosing limits as systemic defences before their implementation. Based on the NICU ME reports, we developed test cases to assess the dosing limits in the NICU infusion pump drug library; the test cases may also apply to other pediatric populations. However, the reliability of test cases could be developed further by using prospective data collection methods, such as direct observation, focus groups and interviews with practitioners to gain even more comprehensive understanding on mechanisms of wrong infusion rate errors within human factors framework [
16,
31,
42‐
44]. Our results indicate that the literature-based calculation formula developed to define the soft upper limits in pediatric intensive care settings [
29] seems to be applicable in NICU settings.
Our results are promising from the perspective of the widely reported risk of alert fatigue associated with poorly defined soft limits [
19,
22]. As expected, the usual dosages did not cause any alerts in this study, while 10-fold errors triggered an alert in all test cases. One of the key factors that made this result possible was the contribution of the neonatologist in a careful assessment of the usual maximum doses of test drugs in collaboration with the research group. Earlier studies have reported clustering of dose error reduction software (DERS) alerts around specific medications and patients (e.g., fentanyl, vasopressin, and insulin in palliative care, when sedatives and analgesics have been significantly escalated) [
21]. Therefore, it would be useful to target similar testing activities to these particular drugs and patient groups as presented in this study.
Our analysis of the ME reports related to wrong infusion rate resulted in similar findings to earlier studies in NICU settings. Most MEs involved a high-alert medication and resulted in overdoses [
1,
2,
7,
11,
12,
32]. MEs can be difficult to identify before reaching the patient because of varying treatment and patient related factors, such as small drug doses and wide size variations between different patients. However, in the NICU settings, the drug library hard limits as system-based barriers have prevented administration of doses even as high as 29-fold compared to the maximum dose [
16,
21,
31]. Especially when high-alert medications are involved, MEs with this size of deviations from the intended dose expose vulnerable NICU patients to serious adverse drug events [
2,
7,
10‐
13,
15,
21]. Following earlier studies, our analysis of contributing factors to wrong infusion rate errors also revealed that failures in the use of other systemic defences or not having them implemented could enable errors [
17]. Consequently, a combination of different preventive error reduction strategies is needed in IV medication use process to mitigate the effects of e.g., environmental, operational and team-work related factors on human performance [
16,
19,
22,
31,
45].
We demonstrated that errors involving doses lower than the usual maximum dose could not have been avoided by using DERS (e.g., the smallest usual doses and most test cases involving a mix-up between two infusion rates). However, a bi-directional smart infusion pump interoperable with the EHR would provide such a solution for even more comprehensive management of human factors contributing to pump-programming errors due to manual adjustment of infusion rate [
16,
18,
28,
31,
46]. The system would enable auto-programming of infusion parameters (e.g., infusion rate) from the EHR system to the pump, which are then verified and followed by starting the infusion by a practitioner [
18]. The pump also automatically sends infusion information (e.g., dose-rate, rate changes, and IV start and stop times) to the EHR system for practitioner confirmation to enable accurate recording of this information in the patient’s record. However, as with smart infusion pumps, the introduction of interoperability with EHR has been associated with challenges, such as inadequate and outdated drug libraries, pump or medications not mapped with the EHR system, and inconsistency in dosing units between the drug library, EHR and usual pump-programming practices [
44].
Our results support the use of weight-based dosing limits in NICU drug libraries, which has been reported as one of the key elements of pediatric drug libraries [
24,
29]. As a result, all the most crucial programming errors (e.g., 10-fold infusion rate) triggered an alert. The test cases related to heparin flush demonstrated that when the medication does not require weight-dependent dosing, the drug library dosing limits are much easier to set. However, it should be noted that when smart pumps are used without EHR interoperability, patient’s weight needs to be entered into the pump when programming the infusion. This represents an additional manual step with a chance for human error [
16,
31].
There are some limitations to the study. First, we used self-reported ME data to create test cases simulating errors resulting in the wrong infusion rate. Self-reporting is associated with the risk of underreporting, and it is unlikely that all errors and near-misses were documented [
47,
48]. The number of ME reports included in qualitative content analysis remained low, as we focused only on one part of the medication use process, and neonates are a limited patient group. However, our aim was to study the possible error mechanisms contributing to wrong infusion rates, specifically in NICU settings instead of error incidence. Therefore, the self-reported ME data was found useful for the purpose of this study. To improve the reliability, two researchers independently searched ME reports meeting the inclusion criteria and verified the findings of the qualitative content analysis, followed by a careful review of the error mechanisms and test cases by the research group, neonatologist, and neonatal nurse practitioners. Nonetheless, qualitative content analysis is a researcher’s subjective interpretation. Some ME reports described the incidents only briefly, so the researchers’ interpretations might not entirely correspond to the actual incidents [
35]. In future studies, the test cases should be further developed by using data collected through prospective methods and other theoretical frameworks, such as focus groups and SEIPS (Systems Engineering Initiative for Patient Safety) [
43,
49].
Second, we only used soft upper limits even though an effective DERS should include hard and soft upper and lower dosing limits [
18,
19,
22]. Earlier studies have reported a high override rate of soft limits, and therefore, all alerts triggered in our study cannot be equated as averted errors in clinical situations. However, not all pump-programming errors cause significant patient harm, which was found out in our ME analysis and has also been observed elsewhere [
25]. Moreover, the number of medications selected to perform the test cases was relatively small, and the selection of different test drugs might have resulted in different findings. When it comes to demonstrating mix-ups between two drug’s infusion rates, the future studies should include designs enabling a more comprehensive exploration of environmental and team-work related factors (e.g., a simulation study with full patient scenarios and multiple end-user participants) [
16,
31,
45].
The current study represents a preliminary work aiming to define dosing limits before their implementation, but the true effectiveness of these limits can be reliably evaluated only after implementation. In future studies, the alert log data and drug library compliance should be studied after implementation of dosing limits to confirm whether the limits have a beneficial effect on drug library compliance and soft limit alert overrides [
18,
19,
22]. Also, a simulation study involving patient scenarios, real care teams and simulated care environments would be beneficial to examine the optimal use of both hard and soft limits [
37]. However, the present study provides NICU and possibly other settings with means for targeting optimal dosing limits, as improperly defined hard limits can prevent legitimate actions. In contrast, unsuitable soft limits can cause useless alerts [
19,
22].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.