Search Process
In all, 55 publications were eligible for inclusion. Three studies reported data from two countries [37–39] (and were considered six unique studies) and four [40–43] reported from the same data set as sister articles [34, 44, 45] (each group was considered one unique study) yielding 54 unique studies. A total of 21,799 articles were excluded at the title review stage as they were either duplications or were not related to medication safety topics. Abstracts were then reviewed and articles excluded if they were not thought to report on the causes of MEs. At the full-text examination stage, only studies focusing on the causes of MAEs in hospitals were included. A summary of the search process is shown in Fig. 2.
Study Characteristics
Country of Origin
Twelve (12/54, 22.2 %) unique studies each originated from the UK [34, 37–40, 46–53] and the USA [6, 7, 38, 42, 43, 45, 54–61]. Six (11.1 %) originated from Australia [62–67]; four (7.4 %) from South Africa [68–71]; three (5.6 %) each from New Zealand [72–74] and Germany [37, 39, 41, 44], and two each (3.7 %) from Canada [75, 76] and Malaysia [77, 78]. The remaining studies were each from separate countries: The Netherlands [79], Spain [80], India [81], Taiwan [82], Iran [83], Ethiopia [84], South Korea [85], China [86], Norway [87] and Turkey [88]. Study characteristics are summarised in Table 1.
Table 1 Background information for included studies
Study Setting and Patient Demographics
A total of 20 studies were carried out in teaching hospitals (37.0 %) and 13 in general or unspecified hospitals (24.1 %). Tertiary care hospitals were the setting for four studies [56, 57, 69, 81] and an army medical centre for another [55]. Two studies were set in paediatrics hospitals [65, 74] and three did not report the institutions from which data originated [59, 68, 71]; however, because they reported MAE data from anaesthetists and inpatient settings, it was assumed that they had originated from hospital environments. Eleven studies were carried out in a range of clinical settings that included hospitals of various types.
Seven studies were conducted solely on paediatric units (13.0 %) [6, 55, 59, 65, 74, 78, 88]. Eleven were carried out using only adult patients (20.4 %). The remainder were either carried out with both adult and paediatric patients (n = 8, 14.8 %) [34, 40, 51, 54, 58, 60, 84, 86, 87] or the ages of patients were not specified (n = 28, 51.9 %).
Fourteen (25.9 %) studies were conducted on only one unit and 22 (40.7 %) on two or more units within each institution. Seven studies did not specify how many units per institution were observed but could have been numerous given the sampling strategy used [42, 43, 45, 52, 54, 58, 63, 82, 85]. Nine studies carried out in anaesthesia could also have sourced data from more than one theatre per hospital based on their sampling techniques [68–72, 75, 81, 86, 87]. Two studies did not state how many units were involved [59, 60].
Study Design
All studies utilising the direct observation method (n = 23, 42.6 %) identified MAEs prospectively, with some confirming the error once the medication administration round had finished (n = 8) and others during this activity (n = 10) (five studies did not specify when MAEs were confirmed). One study utilised observation with chart review [84] and another utilised interviews [67] for prospective and retrospective error identification. Those that did not utilise direct observation to identify MAEs employed a variety of error-detection methods; these included prospective self-reporting methods such as log books (n = 2, 3.7 %) [54, 58], error (e.g. incident) reports (n = 3, 5.6 %) [53, 56, 57] and anaesthetic administration forms (n = 4) [69, 72, 86, 87]. Other prospective methods relied on other healthcare staff or researchers to identify errors through routine activity or chart review [53, 56, 57, 62]. Retrospective methods included interviews (n = 7) [51, 52, 59–61, 63, 66], questionnaires/surveys (n = 10) [42, 43, 45, 64, 65, 68, 70, 71, 75, 81, 82, 85] and focus groups [73, 74]. Some of these studies described a validation process to confirm the presence of an error after the observation period [6, 7, 56, 57, 77].
Besides the use of self-report methods to detect data (n = 16), MAEs were most often collected by pharmacists or pharmacy students (n = 19, 35.2 %) or nurses/nurse students (n = 9). Five studies did not specify who collected the data [50, 59, 63, 77, 78] and five utilised combinations of various healthcare professionals [53, 56, 57, 80, 84].
Data on causes of MAEs were generally collected prospectively whilst directly observing staff activity on the unit(s) (n = 18). Some studies combined observation with chart review [84], informal conversations with staff (n = 2) [34, 40, 41, 44] or interviews (n = 4) [67, 77, 78, 88]. Self-report data was another common method to collect MAE causes; staff used daily log books (n = 2) [54, 58], incident forms for each anaesthetic procedure (n = 4) [69, 72, 86, 87] or surveys/questionnaires (n = 10) [42, 43, 45, 64, 65, 68, 70, 71, 75, 81, 82, 85]. Two studies utilised focus groups to collect these data [65, 73]. Various types of interview were used by the remaining studies to report causes data (n = 11) [51–53, 56, 57, 59–63, 66]. Of those using survey methodology, two used open-ended questions to solicit data [42, 43, 45, 85], three used a limited list of contributory factors from which participants could choose [75, 82, 85] and six did not state the type of questioning employed [64, 65, 68, 70, 71, 81]. Excluding self-reporting methods (n = 16), causes data collectors were most often pharmacists (n = 18), or nurses/nursing students (n = 7).
Definition of a Medication Administration Error
The definition of MAEs varied considerably between studies. Twenty-one (38.9 %) studies did not give a formal or working definition. Nine (16.7 %) studies used their own definition without referencing any established criteria. Of those who referenced criteria as either a complete formal definition or to supplement their own (n = 24, 44.4 %), the most commonly used criteria used were those of Allan and Barker [24] and the American Society of Health System Pharmacists (ASHP) [89, 90].
Route of Administration
Fifteen studies (27.8 %) considered only the intravenous route of administration, whereas the majority (n = 32, 59.3 %) studied all routes of administration. Seven studies observed MAEs via a varying number of different administration routes [7, 37, 38, 47, 55]. Nine studies in the intravenous group involved administration of medication used for anaesthesia [68–72, 75, 81, 86, 87].
Staff Group
The majority of studies investigated errors directly involving nurses (n = 35, 59.3 %), student nurses [63] or both nurses and students [82] or nurses and nurse assistants [80]. Five studies involved nurses and medical staff [34, 39, 40, 77, 78] and eight studies obtained error data from various grades of physician responsible for administering anaesthetics [68–72, 75, 81, 86]. One study involved nurses and anaesthetists in theatres [87]. Two studies obtained data from various healthcare professionals who either made or were directly involved with nursing MAEs [56, 57].
Quality Assessment
Relevance of studies to review question. Overall, few studies were predominantly concerned with the causes of only MAEs (n = 6) [34, 40–45, 52, 63, 82]; most considered these issues after other major objectives such as the prevalence and nature of MAEs or more general experiences of healthcare staff when making errors. Of the latter group, examples included studies that focused on the causes of MEs made by a variety of healthcare professionals [53, 56, 57, 62], and investigations of nurse attitudes towards the defining or reporting of these errors and/or their impact on professional practice [51, 53, 60, 64, 67]. Two asked participants to describe what factors influence their ability to carry out safe practice [73] or medicines management activities [66]. Sixteen studies (29.6 %) did not report any intention to study the causes of specific MAEs.
Sampling. The majority of studies provided insufficient detail of their sampling strategy to determine its nature. A minority (n = 5) reported random sampling of participants [42, 43, 45, 54, 57, 58, 61], with only two describing the method of randomisation [42, 43, 45, 57]. Some publications reported sampling techniques where specific institutions or units were chosen; examples included wards with high error risk [47, 78] or wards chosen to reflect the patient population [76]. Another study recruited nurses from all units within their hospital as part of representative sampling [74]. Some studies interviewed staff based on errors previously identified by other staff members or the researchers [52, 53, 62, 77, 78, 88]. Two papers used the snowball sampling technique to recruit nurse participants [82, 85], two convenience sampling [39, 73] and nine self-reporting based studies sampled the entire population within specified limits (e.g. through registration databases) [68–72, 75, 81, 86, 87]. One study used patients admitted over a specific time period as the sample [56].
Reported sample size varied depending on the study method; 30 (55.6 %) reported the number of staff responsible for drug administration who took part in their study (e.g. nurses). Participant numbers varied between seven nurses in two studies [61, 62] and 720 anaesthetists in another [68]. Of the remaining studies, 21 provided details of the number of errors or the error rate, and three studies did not specify sample size [59, 60, 67].
Causes data collection method. Studies that utilised predominantly quantitative short answer surveys/questionnaires or direct observation methods alone were able to identify important causes of errors; these were generally limited in number (with a few exceptions [64, 82, 85]), did not contain more detail explaining why these causes arose and were not able to specify if multiple causes combined, as data were not generally related to individual errors. These studies listed causes in tables/text using very brief descriptors [6, 7, 38, 39, 50, 55, 64, 65, 72, 76, 79–83, 85]. For direct observation, these results are not surprising considering that MAE causation data from observers that involved opinions or generalisations were excluded (many included such data [6, 7, 38, 48, 50, 55, 76, 80, 83]), leaving only those activities that were factual (e.g. delayed delivery of medication from pharmacy).
In contrast, interviews/conversations (±direct observation), focus groups or self-reporting methods involving narrative free text responses generally provided a greater variety of MAE causes. Some demonstrated the link between administration errors/violations and their associated error-producing conditions using human error theory [34, 40, 41, 44, 53, 62, 88]. Additional verbatim quotes were used to confirm and expand upon data [34, 40–45, 51, 53, 54, 58, 62, 63, 74, 88], with some providing verbatim quotes of individual errors that demonstrated how multiple contributory factors combined to create MAEs [42, 43, 45, 54, 58, 62, 63, 73]. However, not all of these studies provided much information specific to MAEs [53, 60–62, 66, 74], and some provided only brief tabulated/textual data of causes following interviews [51, 52, 56, 57, 61] or direct observation with interviews [77, 78], much in the same way as survey/direct observation studies. Despite this, it is important to recognise that in some cases administration errors were not the sole ME of interest [53, 56, 57, 62], and many qualitative studies did not consider MAEs as their primary research topic.
One study used Reason’s model of accident causation explicitly [53]. Seven studies used criteria that appeared to be based, at least in part, on elements of the systems approach to analysis of errors [6, 7, 55–57, 62, 83]. Three studies used elements of Reason’s model along with other protocols for analysing adverse events [34, 40, 41, 44, 88]. Two studies used other referenced frameworks [64, 80]. Nine studies recorded a single reason or proximal cause for each reported MAE without offering further supplementary detail [6, 7, 38, 50, 56, 77, 78, 80]. Five studies only investigated the causes of clinically significant MAEs [6, 7, 56, 57, 62], with the remainder either basing their data on all types of MAEs or not distinguishing which type they considered (e.g. referred to only as MEs).
A number of causes/factors studies reported methods that had been tested in pilot/pre test phases (n = 15) [6, 7, 46, 51, 54, 58, 62, 65, 70, 77, 78, 82, 84, 85, 88]. Others described run-in [79] or training phases [83] or based their method on earlier work [53, 86]. Only five studies determined causes of MAEs through triangulation of methods [34, 40, 41, 44, 77, 78, 88], which can be used to corroborate findings and, in the case of direct observation research when combined with interviews, bridge the gap between causes of errors that those observing practice cannot identify alone and those who make errors do not notice themselves [25, 40]. Despite this, few of these studies actually reported whether this actually was the case [34, 40, 41, 44].
Reason’s Model of Accident Causation
The data from 54 studies presenting causes data were analysed thematically according to Reason’s model and summarised in Table 2.
Table 2 Summary of medication administration error (MAE) causes reported by included studies
Causes of MAEs
Unsafe Acts
Seven studies reported usable data matching Reason’s description of active failures [34, 40, 41, 44, 53, 56, 57, 62, 88]. The majority of studies identified primary causes of MAEs that could be attributed to the individual responsible for the error without using an established framework. These were broadly considered as either slips, lapses, mistakes or violations.
Slips and lapses. Slips and lapses were common, being identified by 29 studies (53.7 %). Misidentification of either medication or a patient were among the most frequently reported events considered as slips [54, 58, 68, 73, 76, 77, 81]. Misreading either a medication label/product, prescription or other documentation was also common [34, 40, 51, 52, 54, 58, 64, 77, 78]. Staff confused look-a-like or sound-a-like medication names, patient names and medication packaging, which led to MEs [42, 43, 45, 56, 59, 63, 64, 69, 71, 72, 82, 85, 87]. Mental states such as lack of concentration, complacency and carelessness were also reported [51, 52, 78]. Most of the data was presented in tabular or list form in article texts; more detailed examples from qualitative interviews, focus groups or open-ended surveys were able to identify the cause(s) of their slips and lapses [34, 40–45, 53, 62, 63, 73, 88]; these included nurses forgetting to sign a medication order or misreading labels due to being busy and/or distracted [42, 43, 45, 53, 88], failing to administer a drug or being careless due to heavy workload, poor staffing and/or being distracted [42, 43, 45, 73] and selecting the wrong medication due to pressure from others or busy/distracting environments [63].
Knowledge- and rule-based mistakes. Knowledge-based mistakes were less frequent (n = 16), with staff explaining that they did not know enough about the medication they were administering [34, 40, 41, 44, 51, 52, 56, 66, 77, 78, 82], the infusion pump they were using [58] or the patient to whom they were administering it [56, 82]. Rule-based mistakes were generally not observed [88].
Violations. Violations were reported by 14 studies, which were limited predominantly to data collection methods involving (at least in part) conversations with subjects to determine error causality. Where enough data were provided, situational violations (those arising due to necessity [e.g. poorly designed protocols, lack of staff]) were noted [53, 58, 67, 77, 78, 88]. Reasons for violations included trusting senior colleagues [63], patients requests [77], lack of access to suitable administration protocols [88], patient acuity [58, 88], acting in the patients’ interests (e.g. to avoid harm or optimise treatment) [53, 60, 66], poor supervision/drug knowledge (associated with fast bolus intravenous administration) [34, 40], lack of staff (intentionally giving drugs early/late) [58, 67, 78] and common accepted practice (administering without a signed prescription) [53]. One nurse gave paracetamol 4 hourly instead of the prescribed 6-hourly regimen because they thought it would not be effective if given 6 hourly [66].
Other unsafe acts. Calculation errors [39, 48, 50, 52, 64, 74, 77, 78, 80, 87] and faulty checking activities [6, 7, 42, 43, 45, 56, 63, 72, 75, 77, 78, 80–82, 86] were commonly reported. Difficulty with infusion equipment was also noted [42, 43, 45, 56, 58, 76, 83]. Other errors included not following instructions; insufficient evidence existed to determine whether these were deliberate acts (and hence violations) [51, 52, 77].
Latent Conditions
Error- or violation-producing conditions describe the circumstances in which errors occur, and arise due to high-level managerial decisions. Multiple conditions can lead to one unsafe act [36].
The patient. A total of 17 (31.5 %) studies reported patient characteristics as causes of MAEs. Logistical problems associated with delivery of medication were most common and included lack of, difficulty with or delays waiting for intravenous access [34, 39, 40, 49, 50, 67, 68] (leading to wrong route [39], deterioration of medication [49], omission [50], wrong time [49] and compatibility errors [39]), and absent/sleeping patients during drug administration rounds [49, 51, 67, 76, 88]. Severity of patient illness (acuity) was reported by seven studies [42, 43, 45, 54, 55, 82, 86, 88]; some studies provided examples of resulting errors, which included wrong time or dose omission and, in many cases, the nurses were aware of their actions, which would constitute a violation [54, 58, 88]. Patient behaviour also led to MAEs through non-cooperation [34, 40, 51, 85], or prevented errors though knowledge of medications [73].
Policies and procedures. Problems with policies or procedures were reported on few occasions (n = 6). Examples included absence of a policy [41, 44, 56] and policies that were considered over-laborious [42, 43, 45], or generally unsuitable [34, 40, 41, 44, 88] (which led to wrong dose and time violations in one study [88]). Nurses reported that they had only basic information to help them safely mix and administer intravenous medications [41, 44]. Nurses were unclear about the role of the second checker in one study, which contributed to MAEs [53].
Ward-based equipment. Problems with equipment used to aid drug administration contributed to MAEs (n = 19). Insufficient equipment (computers [62] or gloves [78]) [88], malfunctioning equipment [86] and ambiguous equipment design (e.g. syringe driver, drug packaging) [34, 40, 41, 44, 50] were reported; more general problems with drug charts included a lack of access [38, 50] and misplacement [50, 77], which combined with distractions and a noisy environment to lead to a wrong drug error in one interview study [63]. Example(s) of the nature of the infusion pump problems were given by a few studies, which reported that doses could either be administered incorrectly due to being un-calibrated [77] or malfunctioning (a nurse commented how she expected the pump to work because it was well tested) [61], or not be administered at all due to different pump properties [88] or pumps that were not connected [78].
Health and personality. Physical feelings of fatigue, tiredness/sleep deprivation, sickness and general discomfort amongst staff were reported as contributory factors to errors (n = 13) [42, 43, 45, 51, 52, 63–65, 68, 70–72, 82, 86, 88]. More detailed analysis of error accounts by one interview study revealed cases where physical exhaustion was caused by long hours and lack of breaks/food [63]. Staff member mental state at the time of error occurrence was also reported to lead to errors; stress [42, 43, 45, 51, 52, 62, 64], boredom [68], nervousness (with being busy and young) [42, 43, 45] and poor mood [82] were all found to be associated, though their origins were not stated. Personality-related causes were briefly reported as a lack of assertiveness/confidence [52] (including when challenging medical staff [51]), error perception [88] and conscientiousness [51].
Training and experience. Staff inexperience played a role in contributing to errors (n = 8) [34, 40, 42, 43, 45, 52, 72, 82, 86–88]. This included being unfamiliar with the medication, environment, procedures or equipment, as well as being ‘new’ [42, 43, 45, 82, 87]. The feeling of being a newly qualified nurse in post was found in open-ended survey questions to be related to violation-type errors as nurses obeyed/trusted senior colleagues and felt pressure to complete their rounds on time, which led to them not performing their own safety checks [42, 43, 45]. Insufficient training and experience has strong links with knowledge- and rule-based mistakes [34, 40, 41, 44, 53, 88].
Inadequate training was also reported (n = 6), but few specifically mentioned training regarding the practicalities of preparing and administering medication [34, 40, 41, 44, 85]; one study reported that nurses felt their intravenous drug administration skills were not assessed appropriately [34, 40], which may have contributed to nurses learning these skills from each other on the ward [34, 40, 41, 44].
Communication. Difficulty with written communication featured prominently (n = 19), with two studies reporting that illegible [51, 64] and five unclear/messy [34, 40, 41, 44, 55, 77, 82] prescriptions contributed to MAEs. Transcription errors were reported by some [37, 41–45, 50, 76, 83, 85], as were MAEs (e.g. omission and extra dose errors) apparently caused by others’ documentation errors when writing prescriptions or administering medication [42, 43, 45, 50, 60, 65, 67, 77, 78], with open-ended surveys and interviews relating one case to misinterpreting the roles of nursing students and their supervisors [42, 43, 45]. Studies commonly reported more general communication difficulties between healthcare staff or other services without specifying their nature (n = 15). Those using interviews/conversations (± direct observation) and open-ended survey methods reported instances where nurses/doctors failed to pass on information or successfully passed on incorrect information to their colleagues resulting in a drug administration delays [66, 67], drugs being given that should have been withheld [77] and incorrect doses being administered [42, 43, 45, 54]. Problems with labelling were also frequently reported, though detail on their nature and relationship to other causes was missing [49, 68, 70–72, 81, 82, 86].
Supervision and social dynamics. Poor supervision by senior colleagues appears to have a role to play in MAE causation, manifesting as pressuring students to administer drugs more quickly, not supervising or assisting closely enough or giving unclear/incorrect instructions (n = 4) [34, 40, 42, 43, 45, 63, 86]. As discussed previously, specific error examples appear to link poor supervision to violation-type errors and the provoking conditions of inexperience, trusting colleagues and fatigue [34, 40, 42, 43, 45, 63], though supervision has also been linked to poor equipment and workload in one example [42, 43, 45]. Two studies reported apparent overconfidence in/from other nurses when either communicating instructions (as a cause of a wrong dose error) [42, 43, 45] or carrying out independent checks (a dose calculation error) [74]. Pressure from other staff members [42, 43, 45, 71], confronting and intimidating behaviour [64] and social isolation from colleagues also feature as causes [42, 43, 45]. There were examples of how proper supervision and communication could maintain patient safety, through co-workers identifying errors before they reached the patient [54, 63, 73].
Workload and skill mix. Heavy staff workload (n = 19) appeared an important contributor to MAEs, and includes end of shift/patient transfer pressures, patient load and multitasking [34, 40–45, 67]. Resulting errors included omissions [60] and violations [67, 88], though one study found that workload appeared not be a contributory factor (along with most other latent conditions) [57]. Workload was found to combine with distractions to lead to errors in intravenous administration [34, 40] and with patient acuity, inexperience or local working practice to lead to other errors [42, 43, 45, 62].
Skill mix of staff was identified by six studies, with two stating a lack of qualified staff [34, 40, 51] and others that working with inexperienced or new staff members contributed to MAEs [64, 72]. Short staffing was reported by six studies as a cause of MAEs [42, 43, 45, 51, 64, 78, 82].
Distractions and interruptions. Sixteen studies found that interruptions/distractions were a cause of MAEs, though details of the nature of these distractions or their interplay with other contributory factors were rarely given. Of these studies, those using interviews/conversations (± observation) or surveys with open-ended questions provided more descriptive data; examples of resulting errors/near errors included wrong drug [58], wrong time [34, 40] and wrong dose calculations [88]. Distractions included ward rounds [34, 40] or face-to-face/telephone conversations with co-workers/patients [42, 43, 45, 63] and were often present with high workload and/or poor supervision [34, 40, 63].
General work environment. Eleven studies reported on the contribution of the general environment; specifically, noise [64], lighting [64, 71], emergencies [88], and busy [34, 40, 42, 43, 45, 54, 58, 73, 85] or chaotic [42, 43, 45, 54] working environments were identified. Studies offering more detail through open-ended survey questions linked these factors to short staffing, workload, patient acuity and poor supervision [42, 43, 45].
Medicines supply and storage. Issues relating to medicines logistics were reported by 27 studies. A lack of ward stock led to omission/wrong time errors [37, 38, 47, 51, 78–80, 84]. Medication was misplaced or lost on the ward on occasions [47, 62, 75, 81]. In contrast, one study found no errors relating to medication unavailability [50]. The pharmacy department contributed to errors and violations in other cases, through delayed deliveries [50, 55, 66, 76, 85, 88], incorrect dispensing [6, 38, 42, 43, 45, 51, 55, 58] and unavailable stock [58, 78, 79].
Local working culture. Nurses passed on bad practices (e.g. administering without a prescription) that led to errors in three studies [34, 40, 41, 44]. Levels of trust between colleagues [53] and working double shifts or not taking breaks (leading to exhaustion) [42, 43, 45] were additional causes. One nurse described how a wrong drug was selected for administration in a busy and chaotic theatre environment when she/he “relied on routine” [42, 43, 45].
Organisational (high-level) decisions. Included studies rarely reported organisational/high-level decisions as having a direct impact on error occurrence; feedback on errors was considered important by some interview or survey studies using narrative responses to minimise errors in future, and the importance of nurse input in the process was highlighted in one case [42, 43, 45]. Some described supervisory teams responding to errors poorly [63], that opportunities to learn from mistakes were limited [42, 43, 45, 53] and how positive feedback about errors improved nursing practice [63, 64]. More direct causal evidence cited a lack of hospital policy (when challenging other healthcare staff) or misguided policy (low nurse staffing) as causes of MAEs [56, 82], as well as decisions regarding logistical strategy generally revolving around clashes of other ward activities with medication administration [42, 43, 45, 50, 78]. Mix-ups involving medicines that look or sound alike may have roots beyond hospitals with the pharmaceutical industry [34, 40].