Almost every healthcare professional has at some time made a mistake resulting in injury or possible injury to a patient. In the industrial sector it has been acknowledged that human errors will occur, and therefore systems are designed in such a way that errors are prevented or detected before they develop into a true accident. In clinical practice, however, the magnitude of this problem has long been underestimated, despite several large studies confirming the occurrence of medical error with (possible) patient harm.1 In 1991, the Harvard Medical Practice Study reported that adverse events occurred in 3.7% of acute-care hospital admissions in New York. Subsequent analysis showed that more than 50% were caused by errors.2 3 Yet, only since the release of report, “To err is human”, by the Institute of Medicine (IOM) in 1999, has the importance of good patient safety management been recognised by healthcare workers all over the world.

The IOM defines safety as “freedom from accidental injury”. Their report mentioned that the problem of accidental injury is serious, and that patient safety must become a national priority. It was emphasised that the cause is not careless people but faulty systems.4 The IOM recommended that all healthcare settings should establish comprehensive patient safety programmes executed by trained personnel within a culture of safety, and emphasised that reporting systems are one of the key strategies for learning from errors and for monitoring progress in the prevention of their recurrence.5

A good internal reporting system makes all responsible healthcare workers aware of the major hazards, and external reporting allows lessons to be shared so that others can avoid making the same mistakes. Also, external reporting systems will yield a larger sample size, increasing the potential to identify patterns of infrequent, yet striking errors.6 7 According to Leape, most of the benefits can be obtained with specialty based or system-wide reporting programmes.7

Data exist for specialty based reporting systems. The Australian Incident Monitoring System in Anaesthesia used voluntary, anonymous incident reporting, which elicited large volumes of specific information about incidents that could be analysed for root causes and contributing factors.8 9 Through systematic analysis of incident reports, this study, and several others, provided important information about factors associated with adverse events in their specialty.8 10^–12 Patients who are more severely ill, those who are subjected to multiple interventions, and those who remain in hospital longer seem to be more likely to receive a serious injury as a result of medical mistakes.9 13

Thus, newborns in the neonatal intensive care unit (NICU) are a particularly vulnerable group, owing to their small size, physiological immaturity and limited compensatory abilities.14 In a study on medication errors and adverse drug events (ADEs) in two academic paediatric hospitals, the rate of potential ADEs was considerably higher in neonates than in other age groups. Moreover, neonates in the NICU had significantly higher medication error and potential ADE rates than neonates in other wards.15 In establishing a specialty based incident reporting system for NICUs in The Netherlands, the objective was to examine the characteristics of incident reporting systems in NICUs in relation to type, aetiology, outcome and preventability of incidents.

METHODS

Search strategy

Two authors (CS and RAvL) independently searched Medline on PubMed for English, Dutch, German, or French language articles (January 1980 to January 2006). Figure 1 outlines the search strategy. Subsequently, Embase and the Cochrane library were searched with similar search terms as in Medline. The titles or abstracts, or both, of all identified studies were reviewed and full manuscripts obtained for articles meeting the selection criteria (n = 35, fig 1).14 16–49 The reference lists of the articles included and of the review articles were checked manually to look for other potentially relevant articles.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Search strategy.

Analysis of search results

The same authors (CS and RAvL) independently assessed each article for eligibility according to the inclusion criteria (fig 1), and extracted relevant information on study design and results using a predefined data abstraction form. The definition of an incident or error was recorded according to the original author’s definition. If possible, the total number of incident reports was expressed per 100 admissions and per 1000 patient days. Differences between reviewers were discussed within the research group, and agreement was reached by consensus.

RESULTS

Search results

The Medline search yielded no relevant systematic reviews or randomised controlled trials. Eight prospective and two retrospective studies were included (table 1).14 15 23 24 39 41 43 47 50 51

The authors search results were identical, except for two articles.14 30 The authors agreed to include a retrospective study on reported complications14 and to exclude a study concerning proactive risk analysis of the medication-use process.30 Three relevant articles were added after reference checking.15 50 51 An article describing errant drug orders collected by pharmacist review and not by incident reporting was also included because of its relevance.50 A search in Embase and the Cochrane library yielded no additional relevant articles.

Characteristics of incident reporting systems

In most studies, all staff were encouraged to fill out a report on awareness of an error or incident. (table 2).

Most of these studies mentioned the benefits of a voluntary, non-punitive approach to error.15 23 24 39 41 43 47 One study reported a fourfold to sixfold annual increase in the reporting of medication incidents after efforts to reduce the punitive aspect of reporting incidents and to use the educational value of these reports through teaching sessions for nurses.47 Many studies used anonymous reporting to ensure confidentiality. However, because of the anonymity, people could not be contacted for details of reported events.43 In the study by Suresh and colleagues, multi-institutional reporting identified rare but important errors, and error patterns that were unique to the specialty—such as infants being fed breast milk from the wrong mother. This study also mentioned meetings, conference calls and email discussion lists for collaborative learning and systems thinking.43

Type and aetiology

Seven studies reported only medication related incidents or errors.15 24 39 41 47 50 51 Two studies concerned different types of incidents and errors.23 43 Overall, however, medication errors were reported most frequently. Although definitions of an incident or error varied across studies (table 2), data on medication errors in the NICU showed that the total medication error rate was much higher in studies using voluntary reporting than in studies using mandatory reporting (13–14.7 per 1000 NICU patient days, compared with 0.97 per 1000 NICU patient days; and 13–91 per 100 NICU admissions, compared with 0.83 per 100 NICU admissions)(table 3).15 41 47 51

Despite these differences in reporting climate and error rates, almost all authors reported a considerable number of dosing errors due to wrong prescription or wrong administration. When a system approach was used,52 many contributing factors were identified. In one study, up to eight factors were selected for each report.43 Important aetiological factors were failure to follow procedures, inattention and poor documentation or communication (table 3).

Outcome

Although there were many differences in degree and definition of harm across the studies, several conclusions can be drawn. In only one study was no actual harm reported during the study period.50 Potentially harmful incidents, on the other hand, were reported in almost every study. In studies on medication errors, fatal or life-threatening incidents were not often reported. However, in a study describing critical incidents, the need for therapeutic intervention specific to the NICU was 30%(table 3). In this study, incidents relating to ventilation were the most severe and incidents involving drugs the least severe events.23

Preventability

Two studies described the implementation of system changes.24 39 One study found that 63% of the incidents which triggered system changes were classified as minor. System changes involved a standardised prescription form and a computerised system for ordering infant formula bottles.24 In the other study, variations in both the error rate and distribution of error type led to the identification of the cause of wrong-time errors, and subsequently, to the reinstitution of satellite pharmacies near the NICU and PICU.39

Two other studies described the potential impact of prevention strategies.15 51 Kaushal and colleagues reported five preventable ADEs (0.52 per 100 admissions).15 Errors associated with these five incidents included two overdoses, one missing dose, one drug administration error, and giving medication to a patient with a known allergy to this drug. Physician reviewers judged that 93% of the potential ADEs were preventable by physician computer order entry with clinical decision support, and 94% by ward based clinical pharmacists.15 Comparable medication prescription or preparation errors were reported by several other studies.24 39 47 50 Ross and colleagues implemented preconstituted syringes and the double checking of content and administration, after the occurrence of errors caused by confusion of ampoules of different strengths.51

However, only two studies evaluated the impact of preventive strategies.41 50 Results of the first study indicated that paediatric pharmacists could detect errant medication orders and prevent medical errors.50 The second study reported similar results: after the introduction of a pharmacist-led review of medication orders and the introduction of system changes based on incident analysis, medication errors fell significantly from 24.1 (1.7) per 1000 neonatal activity days in the first 4 months and to 5.1 (3.6) per 1000 in the next 3 months.41 However, as these authors pointed out, within the context of their overall risk management approach it is difficult to quantify the proportion of errors reduced by any one change in practice.

DISCUSSION

This review shows that there are few studies of incident reporting systems in neonatal intensive care. Moreover, these studies have limited generalisability, and it is difficult to compare categories used across their systems. However, several conclusions can be drawn from the available data.

Most incident reporting systems in neonatal intensive care use a voluntary, non-punitive approach to incidents. The available data on errors in the NICU suggest that these reporting systems elicit many more incidents in the NICU than a mandatory system, yielding more information in a shorter period of time.

Medication incidents are most often reported in the NICU. This is in concurrence with the literature on incidents in other specialties.13 Fatal or life-threatening harm due to medication incidents was not often reported.23 However, most studies reported that the potential for patient harm as a result of an incident was a significant problem. Moreover, when a system approach was used, many contributing factors were identified. These data suggest that with the use of these incident reporting systems, repeated occurrence of incidents and contributing factors can be identified, thus facilitating their clarification and preventing their recurrence.

However, there are also several limitations to incident reporting systems.

First, owing to anonymity of reports, people cannot be contacted for details of reported events.43 On the other hand, the success of non-anonymous reporting of incidents strongly depends on the creation of a non-punitive climate which allows staff to report incidents without disciplinary sanctions.53

Second, when using a voluntary incident reporting system, only a fraction of incidents may be detected. In an NICU, more complex errors, such as prioritisation of clinical tasks or failure to perform diagnostic assessments, might also result in adverse outcomes. These errors are difficult to measure and, therefore, are not often reported.42 Collaborative learning from incident reports through a neonatal network is likely to offer a solution to this problem.43 However, evaluation of the incidence of actual errors by additional detection methods may be a more accurate marker.15 54 These detection methods can be used for incidents that are likely to be under-reported in voluntary reporting systems.

Finally, assessing the impact of preventive strategies on patient safety remains a challenge. Although voluntary reporting systems increase the number of reports, under-reporting is still likely. Also, a system change can increase the number of incident reports. Therefore, the impact of incident reporting systems and consecutive system changes on patient safety remains a subject for our future investigations.

CONCLUSIONS

Multi-institutional, voluntary, non-punitive, system based incident reporting is likely to generate valuable information on type, aetiology, outcome and preventability of incidents in the NICU. However, the beneficial effects of incident reporting systems and consecutive system changes on patient safety are difficult to assess from the available evidence and therefore remain to be investigated.