Volume 63, Issue 4 p. 340-346
Free Access

Safety in anaesthesia: a study of 12 606 reported incidents from the UK National Reporting and Learning System

K. Catchpole

K. Catchpole

Senior Post-doctoral Scientist, Nuffield Department of Surgery, University of Oxford, Oxford, UK

Search for more papers by this author
M. D. D. Bell

M. D. D. Bell

Consultant in Intensive Care and Anaesthesia, The General Infirmary at Leeds, Great George Street, Leeds, UK

Search for more papers by this author
S. Johnson

S. Johnson

Safer Practice Lead, Department of Safer Practice, National Patient Safety Agency, London, UK

Search for more papers by this author
First published: 10 March 2008
Citations: 78
Dr Ken Catchpole
[email protected]


The incident reporting database at the National Patient Safety Agency was interrogated on the nature, frequency and severity of incidents related to anaesthesia. Of 12 606 reports over a 2-year period, 2842 (22.5%) resulted in little harm or a moderate degree of harm, and 269 (2.1%) resulted in severe harm or death, with procedure or treatment problems generating the highest risk. One thousand and thirty-five incidents (8%) related to pre-operative assessment, with harm occurring in 275 (26.6%), and 552 (4.4%) related to epidural anaesthesia, with harm reported in 198 (35.9%). Fifty-eight occurrences of anaesthetic awareness were also examined. This preliminary analysis is not authoritative enough to warrant widespread changes of practice, but justifies future collaborative approaches to reduce the potential for harm and improve the submission, collection and analysis of incident reports. Practitioners, departments and professional bodies should consider how the information can be used to promote patient safety and their own defensibility.

The high incidence of adverse events in healthcare [1], coupled with evidence that similar errors often recur due to failure of dissemination and remedial action [2], triggered the formation of the National Patient Safety Agency (NPSA) in 2001. This organisation was charged with formulating solutions for existing problems and developing the National Reporting and Learning System (NRLS) for prospective collection and analysis of critical incident reports from healthcare staff, patients or the general public [3]. Although incident reporting is long established within certain specialties, particularly anaesthesia [4], the NRLS, a bespoke multispecialty UK system costing over £5 m ($10.3 m; €7 m) to develop internally and interface with commercial risk management software [3], is the first comprehensive national reporting system for patient safety incidents anywhere in the world.

It is increasingly accepted that adverse outcomes are often due to system failures, whereby deficiencies at many different levels create the context in which human error can have a negative impact [5]. Whilst such serious outcomes should be rapidly disseminated and targeted for remedial action, they may be relatively infrequent even within a large organisation such as the NHS, and are greatly outnumbered by system failures without adverse outcome. The Australian Incident Monitoring System (AIMS), composed of a generic structure and additional systems owned by specialists in several areas (including anaesthetics), has led to a number of important developments in the understanding of safety and in the development of professional practice [6, 7]. It has also been successful in prioritising interventions, developing an awareness of error, and reducing blame, with systems analysis a second-tier benefit [8]. The NRLS similarly encourages blame-free submission of incident reports, with the aim of pre-emptively identifying such deficiencies before harm is generated. The data, covering multiple fields from multiple sources, should help to identify endemic deficiencies in the UK healthcare system, and encourage greater awareness of error and a higher standard of incident reporting and analysis.

In this study, we examined the NRLS database in order to understand the contribution that a larger and more comprehensive database might make to identifying incidents and systemic problems relating to anaesthesia. As well as examining all anaesthetic incidents, we focused on pre-surgery assessment, epidural anaesthesia, and incidents of awareness during anaesthesia, identified by multiple sources including the NPSA’s patient liaison group as numerically common incidents [9–13] that were associated with significant harm. As they cover a broad cross-section of process, treatment and outcome in anaesthetic care, while being specific enough to allow detailed analysis and the identification of underlying systemic deficiencies, these incidents furthermore required us to consider methods for the collection and evaluation of large sets of safety data. This process has consequently informed the development and future use of the broader NRLS.


The NRLS database was examined for all reports relating to anaesthesia from January 2004 to February 2006. Each NLRS report referred to an unintended or unexpected incident that could have or did lead to harm for one or more patients receiving NHS funded care. When an incident report is made and a record of it is stored digitally in a safety management system in a Trust, that information is gathered, de-identified and stored in the NLRS. Incidents can be directly reported to the NPSA via a web-based open-access system [14], but they are usually entered on to a local database by administrative staff or the local clinical risk manager, transcribing a paper-based primary critical-incident report. In order to integrate the wide variety of safety management systems and purposes, the NRLS has 75 data fields, including incident categories at two levels, specialty and location of the incident, and free-text descriptions of the events.

At the time of this analysis, there were approximately 550 000 incident reports within the database. Keyword searches were used to identify incidents of different types (anaesthesia; epidural; assessment; awareness) based on an understanding of how problems might manifest themselves in a reported event (Table 1). A combination of specialty and failure location classifications were then used to exclude cases that were not directly relevant. Finally, removal of duplicate incidents and inspection of the free-text reports confirmed the relevance of the incident events for inclusion in the analysis. However, to maintain the power of a wide sample and ensure all relevant incidents were considered, the keywords used were based on broad definitions. Pre-surgery assessment encompassed both pre-assessment (conducted before the day of surgery), and pre-operative assessment (conducted on the eve or day of surgery). As the assessment process has been developed to identify potential problems before induction of anaesthesia, such as the difficult airway, untreated hypertension, antiplatelet therapy or allergies [15], these were also considered as part of this analysis. Similarly, whilst a traditional approach to awareness would consider general anaesthesia only, pain incidents under regional anaesthesia were included in the analysis of awareness incidents since they were often reported as awareness, and may share similar causal features as well as sequelae such as post-traumatic stress disorder (PTSD) [16–19].

Table 1. Results of keyword searches conducted in the National Reporting and Learning System.
Topic of interest Keyword search Hits Comment
All reported anaesthetic incidents Anaest* OR anest* 12 649 43 disregarded as duplicates
Pre-surgery assessment Anaes* AND assess* 1155 828 incidents selected for further analysis
Anaes* AND unexpected 110 43 incidents selected for further analysis
Anaes* AND Allergy 187 161 relevant incidents
Epidural incidents Epidural 2009 516 selected for further analysis
Epidural AND headache 30 27 selected for further analysis
Epidural AND puncture 16 8 selected for further analysis
Epidural AND abscess 2 1 suitable; 1 added from analysis elsewhere
Awareness during anaesthesia Anaes* AND aware* 59 No relevant cases
Anaes* AND pain 609 18 relevant incidents; 40 other incidents of interest
Anaes* AND awake 2 No relevant incidents
Anaes* AND psycho* None N/A
Anaes* AND PTSD None
Anaes* AND traumatic 12 4 partially relevant incidents; already identified via ‘pain’
  • *Denotes ‘wildcard’ function.
  • PTSD, post-traumatic stress disorder.

In an attempt to identify the most harmful types of incident, we examined simultaneously two data fields (incident category and level of harm). Even in this large dataset, subdividing the data this way potentially yielded only small amounts of data in each category, so for the analysis several harm and incident categories defined in the NRLS were combined. The five levels of harm used in the NRLS were reduced to three levels (no harm; little and moderate harm; severe harm and death). The combination of harm and incident category was then used to estimate the injury effect given the type of incident, in order to account simultaneously for both frequency and severity.


The main findings are summarised in Table 2, which shows the total number of incidents and the outcomes. The most frequently cited incidents – treatment and procedure failures – were examined in more detail using the second level of classification (Table 3). Detailed examination of allergy-related incidents and unexpected incidents are presented in Table 4. Due to their small number, awareness incidents are presented independently in Fig. 1, with 42 incidents associated with failures in the treatment or procedure, seven with equipment failures, four with failures in implementation of care and ongoing monitoring, and three medication failures.

Table 2. Outcomes within categories of incident. Values are number (proportion).
Incident category All anaesthetic incidents Assessment incidents Epidural incidents
n No harm Little or moderate harm Severe harm or death n No harm Little or moderate harm Severe harm or death n No harm Little or moderate harm Severe harm or death
Access, admission, transfer, discharge 630 (100%) 457 (72.5%) 154 (24.2%) 19 (3.0%) 41 (100%) 29 (70.7%) 12 (29.3%) 0 34 (100%) 15 (44.1%) 15 (44.1%) 4 (11.8%)
Clinical assessment (including diagnosis, scans, tests, assessments) 524 (100%) 397 (75.8%) 111 (21.2%) 16 (3.1%) 445 (100%) 344 (77.3%) 92 (20.7%) 9 (2.0%) 14 (100%) 9 (64.3%) 5 (35.7%) 0
Consent, communication, confidentiality 1533 (100%) 1302 (84.9%) 216 (14.1%) 15 (1.0%) 118 (100%) 98 (83.1%) 19 (16.1%) 1 (0.8%) 44 (100%) 31 (70.5%) 13 (29.5%) 0
Documentation (including records, identification) 726 (100%) 656 (90.4%) 69 (9.5%) 1 (0.1%) 66 (100%) 56 (84.8%) 10 (15.2%) 0 10 (100%) 10 (100.0%) 0 0
Implementation of care and ongoing monitoring/review 431 (100%) 269 (62.4%) 141 (32.7%) 21 (4.9%) 35 (100%) 22 (62.9%) 11 (31.4%) 2 (5.7%) 24 (100%) 11 (45.8%) 12 (50.0%) 1 (4.2%)
Infection control incident 274 (100%) 240 (87.6%) 32 (11.7%) 2 (0.7%) 0 0 0 0 8 (100%) 3 (37.5%) 5 (62.5%) 0
Infrastructure (including staffing, facilities, environment) 1123 (100%) 977 (87.0%) 133 (11.8%) 13 (1.2%) 18 (100%) 17 (94.4%) 1 (5.6%) 0 27 (100%) 22 (81.5%) 5 (18.5%) 0
Medical device/equipment 1664 (100%) 1348 (81.0%) 269 (17.8%) 20 (1.2%) 19 (100%) 15 (78.9%) 4 (21.1%) 0 57 (100%) 44 (77.2%) 12 (21.1%) 1 (1.8%)
Medication 1120 (100%) 880 (78.6%) 225 (20.1%) 15 (1.3%) 47 (100%) 34 (72.3%) 13 (27.7%) 0 16 (100%) 7 (43.8%) 8 (50.0%) 1 (6.3%)
Patient accident 510 (100%) 286 (56.1%) 211 (41.4%) 13 (2.5%) 12 (100%) 4 (33.3%) 8 (66.7%) 0 85 (100%) 63 (74.1%) 22 (25.9%) 0
Treatment, procedure 3839 (100%) 2561 (66.7%) 1166 (30.4%) 112 (2.9%) 212 (100%) 125 (59.0%) 73 (34.4%) 6.6% (14) 222 (100%) 131 (59.0%) 86 (38.7%) 5 (2.3%)
Other 232 (100%) 122 (52.6%) 88 (47.9%) 22 (9.5%) 34 (100%) 20 (58.8%) 12 (35.3%) 2 (5.9%) 11 (100%) 8 (72.7%) 2 (18.2%) 1 (9.1%)
Total 12606 (100%) 9531 (75.3%) 2849 (22.5%) 269 (2.1%) 1035 (100%) 760 (73.4%) 247 (23.9%) 28 (2.7%) 552 (100%) 354 (64.1%) 185 (33.5%) 13 (2.4%)
Table 3. Level-2 classification of treatment and procedure incidents identified in the National Reporting and Learning System, with outcomes. Values are number (proportion).
Incident category level 2 Total No harm Little or moderate harm Severe harm or death
Delay/difficulty in obtaining clinical assistance 145 (100%) 113 (77.9%) 30 (20.7%) 2 (1.4%)
Delay/failure in recognising complication of treatment 55 (100%) 31 (56.4%) 19 (34.5%) 5 (9.1%)
Delay or failure to monitor 96 (100%) 63 (65.6%) 23 (24.0%) 10 (10.4%)
Documentation – missing/inadequate/ illegible healthcare record/card 40 (100%) 32 (80.0%) 7 (17.5%) 1 (2.5%)
Documentation – no access to 5 (100%) 5 (100.0%) 0 0
Extended stay/episode of care 28 (100%) 24 (85.7%) 3 (10.7%) 1 (3.6%)
Failure to discontinue treatment 3 (100%) 3 (100.0%) 0 0
Failure to follow up missed appointment 4 (100%) 4 (100.0%) 0 0
Inappropriate patient handling/ positioning 60 (100%) 27 (45.0%) 33 (55.0%) 0
Infusion injury (extravasation) 30 (100%) 10 (33.3%) 20 (66.7%) 0
Missing needle/swab/instrument 62 (100%) 58 (93.5%) 4 (6.5%) 0
Other 1332 (100%) 687 (51.6%) 588 (44.1%) 57 (4.3%)
Patient incorrectly identified 31 (100%) 28 (90.3%) 3 (9.7%) 0
Retained needle/swab/instrument 20 (100%) 17 (85.0%) 3 (15.0%) 0
Theatre list details incorrect 54 (100%) 50 (92.6%) 4 (7.4%) 0
Transfer – delay/failure 7 (100%) 6 (85.7%) 1 (14.3%) 0
Treatment/procedure – delay/failure 1283 (100%) 954 (74.4%) 306 (23.9%) 23 (1.8%)
Treatment/procedure – inappropriate/ wrong 481 (100%) 378 (78.6%) 96 (20.0%) 7 (1.5%)
Treatment not clinically indicated 9 (100%) 8 (88.9%) 1 (11.1%) 0
Unplanned return to theatre 94 (100%) 63 (67.0%) 25 (26.6%) 6 (6.4%)
Table 4. Failures in the pre-surgery assessment process, reported to the National Reporting and Learning System.
Allergy incidents Incorrect identification of allergies Patients with allergies that were not identified
Patients identified as allergic but were not
Failure to appropriately communicate the allergy to all team members Missing or hidden allergy bands Allergies omitted or recorded in the wrong part of the documentation
Allergies not communicated to the right people at the right time
Inappropriate treatment on the day of surgery Poor preparation for allergic patients Violations of policy
Pressure not to cancel list following a late identification of allergy
Unexpected events Unexpected physiological reactions Cardiac problems
Loss of blood pressure/haemorrhage
Unexpectedly long recovery from anaesthesia
Unexplained deaths
Medication problems Patient received insufficient anaesthetic
Epileptic seizures under anaesthesia
Reaction of patient to medication
Seizure following medication
Communication and management problems Delay of blood products Last minute list changes meaning a different patient was expected in theatre
Inappropriate transfer requiring unexpected intubation & ventilation
Unexpected/unclerked admission
Patient and team not informed of scan appointment
Failure to inform patient of need to fast
Basic anaesthetic problems Unexpectedly difficult intubation (sometimes resulting in dental damage)
Vomiting leading to aspiration
Displaced tracheostomy tube
Details are in the caption following the image

Incidents classified in the National Reporting and Learning System as awareness during anaesthesia. Solid – general anaesthesia; diagonal shading – regional anaesthesia.


It is acknowledged that the lack of detail inherent in generic data fields prohibits translation of these results into robust arguments for immediate change in clinical practice. However, the large dataset and the range of suggested causal factors that echo the known high potential for patient harm in anaesthesia [20] suggest a need for both reflection and action. Although harm only occurred in about one quarter of the incidents, the number and range of incidents supports the view that the provision of care is complex and risky, and that the failure to prevent harm is not the only evidence of deficient practice. Reducing the opportunity for harm, and enhancing the recovery from harm, will provide a more powerful solution than either approach in isolation [21]. This emphasises the value of learning from both harm and no-harm incidents [22].

It is encouraging that some systemic deficiencies identified here are currently being acknowledged or addressed through the improvement of technologies and procedures [23], information and documentation [12, 24], pre-operative communication [25], and continuity of care [26–30]. The need for training in the recognition and prevention of potentially harmful situations has been recognised for some time [31–35], which may be especially important for epidural care where the response to complications is critical [9, 36–38], and for awareness incidents, where certain scenarios such as emergency surgery run the narrow line between the harms of awareness and cardiovascular compromise and where awareness monitoring technology may not be completely reliable [11, 39, 40]. Such combinations of reduced opportunity for harm, and better recognition of and response to potential harm, may offer significant improvements in quality of care, and the NLRS data are capable of informing that debate.

This study also tells us much about the current limitations of incident reporting systems. The NPSA have encouraged hospital trusts to ensure that staff know what and how to report, and provide support and advice through regional representatives, in an attempt to ensure compliance and consistency. Nonetheless, potentially just one in five incidents are reported [41]. Aside from this, data are subject to classification and hindsight bias [42, 43], with a fundamental difficulty in assigning a single classification to what is often a cascade of errors [44], all compounded by the lack of a universal taxonomy and analysis framework [45]. Medication or treatment failure may be difficult to distinguish, particularly in anaesthetics, and the hierarchy in the taxonomy may confuse or hide incidents, with for example, an extravasation injury (Table 3) also classified as a medication error (Table 2). Timing and location of events can also be difficult to ascertain, meaning that it was difficult to reliably establish, for example, whether epidural complications were post- or intra-operative, limiting their comparison with general anaesthesia. There were also insufficient data to determine the problems associated with different analgesic treatments or techniques, such as patient controlled analgesia, that would be particularly useful for informing the epidural risk-benefit debate.

We did not establish in this study who had reported, or the route by which the report arrived in the NRLS, though it seems reasonable to assume that the reports were most frequently made by nurses [46]. However, data are potentially skewed in the NRLS when reported by parties other than specialists or the practitioners directly involved. For example, the higher levels of harm found in the epidural incidents may reflect a bias away from reporting no-harm events [22] given that many epidural complications may not be considered as incidents [46], with the high number of ‘other’ and ‘treatment/procedure’ classifications likely to be a function of taxonomic limitation and non-specialist reporting, and not purely the frequency of these incidents. The particularly high proportion of deaths rate in the ‘other’ category may also be a reflection of the difficulty in assigning a single cause in more serious and complex situations [44, 47]. Nevertheless, despite all these inconsistencies and the difficulty of establishing reliable data, the results for anaesthesia and the selected subgroups were broadly comparable with data derived from the Clinical Negligence Scheme for Trusts [48].

The free-text analysis, through keyword search and subsequent inspection, offered by far the richest source of data, with broad definitions aimed to capture incidents classified under a wide range of causes. Whereas 524 incidents were classified by the NRLS database as being related to assessment, keyword searches of the free text revealed a total of 1035 incidents where assessment had been implicated. However, with free-text analysis it was not always possible to distinguish between incidents where care in anaesthesia had been directly compromised, or where anaesthesia was involved, but was not of central importance such as with errors in surgery. Until a sufficiently validated reporting and analysis framework becomes available, substantial classification inconsistencies must therefore be assumed.

With specialist involvement, invaluable in achieving improved data quality, efforts to elicit more accurate and detailed reports through structured data processing is proving successful in Australia [49]. Further understanding of error may be possible through a unifying national reporting system such as the NRLS, but if not owned by the specialty, it may not have the necessary refinement of data collection for meaningful interpretation. Individual practitioners, departments, institutions and professional bodies should consider how data of this nature can be used in the future, how further study should be directed and where responsibility lies for addressing the areas of concern, noting that certain solutions require the highest level of collaboration [50]. Critical to the success of this strategy will be a sense of ownership of incident reporting and confidence that the material will be used for learning and service improvements rather than emphasising personal liability.


The authors would like to thank all those who have taken the time to report these incidents, the NPSA for funding and supporting this work, and the editors and reviewers for their substantial efforts to help in the publication of this paper.