OUP user menu

Cardiac surgery errors: results from the UK National Reporting and Learning System

Elizabeth A. Martinez, Andrew Shore, Elizabeth Colantuoni, Kurt Herzer, David A. Thompson, Ayse P. Gurses, Jill A. Marsteller, Laura Bauer, Christine A. Goeschel, Kevin Cleary, Peter J. Pronovost, Julius Cuong Pham
DOI: http://dx.doi.org/10.1093/intqhc/mzq084 151-158 First published online: 11 January 2011


Objective To describe cardiac surgery-related incidents and compare the types and severity of incidents occurring in the operating room (OR) versus non-OR locations. We hypothesized that the type and severity of incidents in cardiac surgery would differ in the OR compared with non-OR locations.

Design A retrospective cross-sectional study of all incidents categorized as cardiac surgery in the UK National Reporting and Learning System database between January 2003 and February 2007. Differences in proportions were evaluated by χ2 or Fischer's exact test. The odds ratio of an event occurring in the OR compared with all non-OR settings was calculated using logistic regression. The harm susceptibility ratio ranked locations by the degree of harm.

Setting All trusts performing cardiac surgery.

Participants None.

Intervention None.

Main Outcome Measures Cardiac surgery incidents occurring in the OR versus non-OR.

Results A total of 4828 (<1%) incidents from 55 trusts were designated as involving cardiac surgery patients during the study period; 21% occurred in the OR. Overall, 32% of incidents resulted in harm: 23% of OR and 34% of non-OR incidents. The distribution of incident type and harmful incidents differed in the OR compared with the non-OR setting (P < 0.05).

Conclusions Our findings offer unique insights into the types of incidents occurring during cardiac surgical care in the UK. In the OR, interventions should focus on reducing errors associated with medical devices/equipment, whereas outside the OR, they may focus on medication errors and patient accidents.

  • cardiac surgery
  • surgical errors
  • perioperative
  • operating room
  • medical errors
  • patient safety
  • error reporting
  • incident reporting


The complexity of cardiac surgery and the patient's acuity exponentially increases the risk for medical errors. In a landmark study, Brennan et al. [1] estimated that up to 3.7% of all US hospital admissions and as many as 7% of general surgery admissions are associated with an adverse event. A study of cardiac surgery patients in Utah and Colorado found an adverse event rate of 12.3% (95% confidence interval [CI], 7.9–18.7%) compared with 3.0% (95% CI, 2.7–3.4%) among all surgical admissions [2]. A prospective single-hospital study assessed four major surgical services and found a complication rate of 26.9% and a mortality rate of 3.34% among cardiothoracic patients [3]. They also determined that 49.5% of minor and 38.7% of major complications and 25% of deaths among cardiothoracic patients were avoidable.

Although cardiac surgery is common and the international focus on surgical errors intensifies, errors remain high, resulting in needless patient morbidity and mortality [3]. Using the conservative, lower bound estimated rate of 7.9% of adverse events [2], 28 000 of the 357 000 patients undergoing a cardiac surgical procedure annually in the USA [4] will experience an adverse event. Approximately half of these events are likely preventable [2, 5].

Mitigating preventable harm is a challenge because of the limited research done in this area. Although there is some evidence on the frequency and impact of errors among cardiac surgery patients, most of the data are over a decade old [2] or come from a single institution [3]. Moreover, these studies did not evaluate adverse events by clinical area or severity of harm, limiting their ability to bring patient safety improvement efforts into focus.

Adverse event reporting systems provide a unique opportunity to review the spectrum of adverse events in cardiac surgery. Although reporting systems cannot provide information about incidence rates, they can provide useful data in understanding the location, types and severity of events. The National Reporting and Learning System (NRLS) in the UK is among the largest national reporting systems with over four million reports, making it a particularly rich source of information. We undertook this research to gather background data for a larger prospective observational interdisciplinary study which was part of the Flawless Operative Cardiovascular Unified Systems initiative, to identify hazards in cardiac surgery and apply methods to reduce harm. This paper describes cardiac surgery errors reported to the NRLS and compares the types and severity of incidents occurring in operating room (OR) versus non-OR locations with the goal of identifying areas in which to target interventions. We hypothesized that the type and severity of errors in cardiac surgery would differ in the OR compared with non-OR locations.


We performed a retrospective cross-sectional study of all incidents categorized as cardiac surgery events in the NRLS database between January 2003 and February 2007. The NRLS is a voluntary web-based national incident reporting system managed by the National Health Service's (NHS) National Patient Safety Agency (NPSA) in the UK. The reporting system has been operational since November 2003 and all NHS organizations in England and Wales were able to submit data by January 2005. Since its inception, over four million incident reports have been submitted. Incidents are defined as any unintended or unexpected event that could have or did lead to harm for a patient receiving NHS healthcare [6].

Reporting and participating in the system is voluntary. More than 60% of NHS trusts (a trust represents a set of hospitals) report at least monthly, with 70% of the reports coming from acute care trusts. Approximately 99% of the incident data collected by the NPSA comes from the local trust's risk management system in either a paper or an electronic format. Once a local risk management system has been populated with incident data, the NPSA maps (links) the incident reports to the national database (this occurs the first time a trust submits data to the reporting system). After this initial mapping, incident data can be submitted from the trusts directly to the reporting system, where it is collected, cleaned, stored and used for analytical purposes. The Quality and Safety Research Group of the Johns Hopkins University performed the analysis reported here under agreement from the NHS. The study was approved by The Johns Hopkins University Institutional Review Board.

Data abstraction

A total of 313 trusts had reported incidents to the NRLS at the time of data abstraction in February 2007, including all trusts performing cardiac surgery; 55 reported cardiac surgery events. We abstracted characteristics of the medical incident such as time of day (day or night defined as 07:30–19:29 and 19:30–07:29, respectively), time of week (weekday or weekend), cause, incident type, work area (described below) and harm level from reports involving cardiac surgery. Harm was reported by the user and classified in the reporting system as no harm, low harm (minimal harm—patient(s) required extra observation or minor treatment), moderate harm (short-term harm—patient(s) required further treatment, or procedure), severe harm (permanent or long-term harm) or death. All data elements except patient age and gender were analyzed at the incident level. Patient age and gender were analyzed at the patient level.

The reporting system used the patient's service at the time of the event to designate the error as cardiac surgery. We categorized cardiac OR errors as ‘cardiac surgery’ occurring in the ‘operating theatre’ or the ‘anesthetic room,’ and cardiac non-OR errors as ‘cardiac surgery’ occurring in the ‘intensive care unit (ICU),’ ‘recovery room,’ ‘ward’ and ‘other’ (hospital buildings [inside], hospital grounds [outside], hospital transport [car], laboratory, pharmacy, radiology and therapy department). There were 12 incident types in our analysis defined by the reporting system (described in Table 2), including ‘other’ (disruptive, aggressive behavior; patient abuse [by staff/third party] and self-harming behavior).

Data analysis

Descriptive statistics were used to summarize incident characteristics. Differences in proportions were evaluated by χ2 test or Fischer's exact test as appropriate; differences in means by t-test. A logistic odds ratio was calculated to determine the odds of an event occurring in the OR compared with non-OR location, accounting for the correlations within trusts. We also performed a descriptive case analysis in a subsample of 83 incidents that resulted in death. Statistical analyses were performed using STATA v.9.2. SE (College Station, TX, USA) and SAS 9.13 (Cary, NC, USA).

To rank the incident locations according to the degree of reported harm (regardless of the incident type), the odds of reported harm for each incident location were compared with the average odds of reported harm across all other incident locations. These statistics are referred to as harm susceptibility ratios [7]. A harm susceptibility ratio of greater/less than 1 for a particular location indicates that incidents attributable to that location had higher/lower reported odds of harm compared with the average odds of reported harm for the cumulative incident types. The harm susceptibility ratios were calculated from a random-effects logistic regression model that accounted for three possible sources of variation in the data: (i) across trusts, (ii) across incident locations and (iii) across incident location within a trust. The normal approximation of the log odds of harm was used. The three level random effects were considered independent and normally distributed with respective variances. Trusts with less than 50 medical error incidents were excluded from the analyses. We also excluded trusts with five or fewer harmful reports and five or fewer non-harmful reports.


Of 983 660 incidents reported from 2003 to 2007 to the NRLS, 4828 (<1%) were designated as cardiac surgery (Table 1). All of the trusts that perform cardiac surgery are represented in this analysis. The overall mean age of patients was 62 years. Twenty-one percent (n = 1004) of incidents occurred in the OR (Table 1). There was no significant difference in the mean age of patients experiencing an incident in the OR compared with non-OR settings. Ninety-two percent of OR errors occurred during the day shift. There were missing data from gender (62%), time of incident (29%) and day of incident (5%). Seventy-seven percent of OR incidents reported did not result in harm compared with 66% of non-OR incidents. Overall, 32% of incidents (n = 1522) resulted in harm: 23% of OR incidents and 34% of non-OR incidents. The harm susceptibility ratios for the two locations were: 0.84 (95% CI = 0.54–1.30) for the OR and 1.21 (0.80–1.85) for the non-OR.

View this table:
Table 1

Characteristics of cardiac surgery errors (n = 4828)

Operating rooma (n = 1004)Non-operating room (n = 3824)
Patient demographics
 Mean age in years (±SE)63 (61–65)61 (60–63)
 Female, no. (%)b79 (30)518 (33)
Event characteristics
 Time of dayb
  Day shift, no. (%)669 (92)c1655 (61)
  Night shift, no. (%)55 (7.6)c1056 (39)
 Harmful incidents per day shift, no. (%)135 (20)c451 (27)
 Harmful incidents per night shift, no. (%)16 (29)32 316 (31)
 Time of the weekb
  Weekday (Monday–Friday), no. (%)913 (95)c2860 (79)
  Weekend (Saturday–Sunday), no. (%)49 (5.1)c744 (21)
Severity of harm (n noted by line item)
 No harm (n = 3306), n (%)772 (77)2534 (66)
 Low/moderate harm (n = 1439), n (%)214 (21)1225 (32)
 Severe harm/death (n = 83), n (%)18 (1.8)65 (1.7)
 Harm susceptibility ratio (95% CI)0.84 (0.54–1.30)1.21 (0.80–1.85)
  • aOperating room includes operating theatre and anesthetic room, as reported directly to the NRLS. bPercent missing data for gender (62%), time of day (29%) and day of week (5%). cP < 0.05 by χ2 test.

Table 2 describes cardiac surgical incidents by type and harm level, comparing the OR with the non-OR setting. The distribution of incident types in the OR differed from those in the non-OR setting (P < 0.0001). The most common type of incident in the OR was related to treatment/procedure (32%) and in the non-OR was patient accident (32%). For each incident type, the odds of an incident occurring in the OR compared with non-OR settings were statistically significantly different, except for infrastructure and the ‘other’ category. The odds ratio of an incident occurring in the OR was 6.0 times higher than the non-OR setting for a medical device/equipment incident, 4.61 times higher for an infection control incident and 3.67 times higher for a treatment/procedure incident.

View this table:
Table 2

Cardiac surgical incidents, by type and location

Incident typeAll incidentsaHarmful incidentsa,b
% operating roomc (n = 1004)% non-operating roomc (n = 3824)Odds ratio of incident in the operating room (95% CI)% operating roomc (n = 232)% operating roomc (n = 1290)Odds ratio of harmful incident in the operating room (95% CI)
Access, admission, transfer, discharge1.55.30.27 (0.11–0.65) (0.57–1.65)
Clinical assessment (including diagnosis, scans, tests, assessments) (0.23–0.94) (0.37–2.30)
Consent, communication, confidentiality7.44.71.60 (1.02–2.53) (1.06–6.64)
Documentation (including records, identification) (1.11–2.39) (1.20–7.54)
Implementation of care and ongoing monitoring/review1.14.00.27 (0.14–0.51) (0.11–0.70)
Infection control incident8.92.14.61 (2.51–8.46) (0.28–2.39)
Infrastructure (including staffing, facilities, environment) (0.91–1.68) (0.98–2.72)
Medical devices/equipment275.86.00 (3.15–11.4)163.45.37 (2.58–11.2)
Medication2.4170.12 (0.07–0.22)2.2120.16 (0.07–0.36)
Patient accident2.1320.05 (0.02–0.09)6.0410.09 (0.03–0.25)
Treatment, procedure32123.67 (1.98–6.80)43154.41 (1.95–9.98)
Otherd0.71.40.50 (0.20–1.25) (0.23–0.72)
  • CI, confidence interval. aDistribution of all incidents and harmful incident types differed by location, P < 0.05. bHarmful incidents include those resulting in low, moderate or severe harm, or death. cOperating room includes operating theatre and anesthetic room, as reported directly to the NRLS. dOther includes disruptive, aggressive behavior; patient abuse (by staff/third party), self-harming behavior; ‘other’ as defined natively in the NRLS.

Of the incidents resulting in harm, there were higher odds of a medical device/equipment (5.37) event, treatment/procedure (4.41) problem, documentation (3.01) issue and a consent/communication/confidentiality (2.65) issue occurring in the OR compared with the non-OR setting (Table 2). Moreover, 15% (n = 232) of incidents resulting in harm occurred in the OR and 85% (n = 1290) occurred in a non-OR setting. Table 3 describes a sampling of the 83 incidents leading to severe harm or death, regardless of location, with a verbatim case report.

View this table:
Table 3

Incidents leading to severe harm or death

Type of incidentn (%), n = 83Case report abstract sample
Access, admission, transfer, discharge7 (8.4)Patient with ischemic heart disease: in transfer from XX Hospital; ongoing chest pain; 90% LMS; good LV; CABG × 2 4.5.05. Bilateral IMA, principal surgeon: XX). Postop cardiac failure and hypoxia Day 1; reintubated, right heart failure on TOE [echo], redo CABG (principal surgeon: XX), prolonged hemostasis. Postop: TOE [echo] acute right side failure, Levitronics RVAD inserted in ITU 6.5.05. Hemofiltered. Outlook: poor. Outcome: awaited. Issue: patient selection for indirect surgical consultant supervision of registrars
Clinical assessment (including diagnosis, scans, tests, assessments)7 (8.4)Patient had a complete heart block following a mitral valve replacement (with a mechanical prosthetic valve). She was paced with a temporary pacemaker and was awaiting a permanent pacemaker implant. She was anticoagulated with Heparin intravenous infusion and an optimum anticoagulation level was aimed for. The blood results showed over—anticoagulation resulting after sometime in a cardiac tamponade requiring an emergency reoperation
Consent, communication, confidentiality3 (3.6)Pt [patient's] daughter alerted staff that pt was not responding. Pt found weak and not breathing, bradycardic heart beat. Crash buzzed and team called. Pt began breathing spontaneously and heart rate increased after 1 min. After crash team arrived it was noted pt was not for resus [resuscitation]. Correct form in note. This was not handover by night staff or on sheet
Documentation (including records, identification)2 (2.4)The group and save in the patient notes was for a [Patient name] O Rh pos. The patient was actually [Patient name 2] O Rh neg
Implementation of care and ongoing monitoring/review5 (6.0)Long-term patient on CTICU ward gained grade 2 pressure ulcer to sacrum despite hourly stands and 2° turns when in bed
Infection control incident7 (8.4)Hosp acquired MRSA blood stream infection
Infrastructure (including staffing, facilities, environment)4 (4.8)Patient was second case on Mr [Staff Name]’s list on Friday. 3 cases in total listed but only 2 ICU beds available due to staffing problems. Decision to cancel operation taken at 14.00
Medical devices/equipment2 (2.4)Patient in cardiac ORs, anesthetized and at end of CABG operation. Coming of bypass—air accidentally went into aortic line. Patient put back on bypass machine for further period of time, at end of surgery patient taken to CITU
Medication16 (19.0)Severe bleeding 3 days after surgery because of wrong heparin dosage
Othera5 (6.0)Patient referred and placed on waiting list on xx. Patient admitted on xx. Patient died during a procedure to have pacing wire inserted on xx
Patient accident8 (9.6)Heard patient fall—he called out. Found pt on bathroom floor. He said he slipped on wet floor while having his morning wash. Was able to get up himself but hit his head on the floor. GCS 15. Multiple skin tear on left back shoulder and elbow. Complaining of right hip pain and back pain. SHO informed patient not reviewed until 13.30
Treatment, procedure17 (20.0)Patient was transferred from the OR to recovery on arrival was extubated, which is an unusual course of treatment at this centre post cardiac surgery. The patient subsequently had a respiratory arrest, and required reintubation. After a brief time of stability post reintubation the patient suffered a cardiac arrest and was taken back to OR for resternotomy. In OR, the surgical team was unable to establish cardiopulmonary bypass and the patient died
  • aOther includes disruptive, aggressive behavior; patient abuse (by staff/third party), self-harming behavior; ‘other’ as defined natively in the NRLS.


This analysis revealed that 21% of reported cardiac surgery errors reported to the NRLS in the UK occurred in the OR environment. This is a high percentage given the brief time a patient spends in the OR. For example, if an average cardiac procedure takes 6 h and the average length of stay is 6 days (144 h), then 21% of the errors occur within the first 4% of a patient's time in the hospital. This is not surprising considering the complexity of procedures and surgical team dynamics in the OR. Human factors research has found that the use of new technology [8], team member personalities [9], communication failures [10], and high workload and competing tasks [11] affected team performance and jeopardized patient safety in the OR. Research findings also suggest that surgical staff may perceive that deviating from safe practices is admissible because there are safety nets to catch errors [12]. Although the majority of errors in this analysis did not result in reported harm, these incidents provide important information. It has been shown that the number of minor events (described as anything that disrupted surgical flow but in isolation was not expected to impact the safety of the patient) decreases the cardiac surgery team's ability to compensate for a future major event in the same case [13, 14]. Thus, no-harm incidents may reflect latent failures in the system that, under different circumstances, could result in harm to the patient or reduce the system's ability to compensate for major errors.

We found a trend toward an increased likelihood of harm based on the harm susceptibility ratio for OR and non-OR locations (bottom of Table 1) as the extent to which patients are monitored decreased. Although this trend did not reach statistical significance, this may suggest an important potential relationship and a possible key to understanding cardiac surgery errors and, importantly, how and where to target interventions. In our data set, the most common cardiac OR incident types were treatment/procedure and medical device/equipment-related, and both were very likely to result in harm should they occur. The next most common event type in the OR was associated with infection control. However, infections were less likely to result in harm in the OR when compared with non-OR settings. It is difficult to interpret this finding because the person reporting the incident scored the level of harm and often unaware of harm that developed latter. It could be that infection control events in the OR incur less harm, but it is equally likely that OR staff did not observe the harm because it developed or was visible only after the patient left the OR. In the case of mediastinitis, for example, patients can be diagnosed with a surgical site infection up to 1 year after their surgery date. Because OR providers who initially cared for the patient rarely get this feedback, they may underestimate the impact of an infection control incident and the degree of harm.

In our study, only 2.4% of OR errors were medication-related compared with 17% in non-OR settings. Previous studies, however, have shown that medication errors during surgical procedures are common and result in significant harm to patients [1519]. In the Australia Incident Monitoring Study, medication-related incidents were the most commonly reported problem and constituted 30% of all reports submitted by anesthesiologists [15, 16]. The OR is, theoretically, a high-risk area for medication errors because OR providers bypass many of the safety checks (e.g. computer support, nurse, pharmacist) used in other areas of the hospital. Without baseline performance, it is difficult to assess whether our finding corresponds to progress in OR medication safety since the Australian study was published. Alternatively, our findings may reflect underreporting of medication events or increased reporting of other event types and may relate to reporting differences by provider type or by local cultures. For example, nurses were the most frequent reporters to this system. If medication errors did occur, however, they were assessed as less likely to cause harm in the OR compared with the non-OR setting. Perhaps, in the case of the OR, an anesthesia provider who is taking care of only one patient may be less likely to commit a medication error compared with the combination of a physician, pharmacist and nurse taking care of numerous patients on the ICU or ward. This would represent a departure from conventional wisdom regarding medication safety and requires additional investigation.

There is a growing body of literature regarding safety in cardiac surgery [3, 9, 2025]. As in our study, errors have been found in all phases of cardiac surgical care. Within the OR, communication and distractions contribute to all types of incidents, especially those that cause harm [2224, 26]. Our analysis adds to this important body of knowledge and can help identify areas in which to focus error reduction efforts.

This study has several limitations. First, the data are self-reported and represent a non-random sample of errors from an unknown universe of errors. Underreporting of errors in voluntary systems is well known [27, 28]. As such, these data cannot generate incidence rates or capture the entire universe (denominator) of errors. Nevertheless, they represent a collection of hazards that staff have recognized and the risks identified are likely real. Second, there are missing data, such as gender; however, these variables were not necessary for our analyses of incident distribution or associated harm. Third, there were no baseline data for the determination of improvement. Fourth, the coding of harm and other variables may have some misclassification. Although all staff received some training on the grading of harm, there is no standardized training throughout the UK. Finally, the results are from 55 trusts in the UK and may not be generalizable outside of the UK.

Despite these limitations, the study has substantial strengths. It uniquely stratified cardiac surgery errors by location and level of harm, offering important information for providers in decreasing the risk of preventable complications in the cardiac surgical patient population. This information must be given back to the providers so they can learn from these errors and improve their performance. To our knowledge, this is the largest known collection of self-reported incidents in cardiac surgery to date, providing a sample that can be used to help inform patient safety improvement. Efforts to improve safety can focus on the most common incidents in each location, on those that result in the highest likelihood of harm, or some combination of the most common and severe. To prioritize and use resources wisely, efforts may focus on evaluating medical equipment/device use (e.g. is the equipment easy or difficult to operate, is the human–machine interface problematic, what are the current training and certification policies for clinicians, are the equipment safety standards being maintained, is the equipment being used correctly) in the OR because these events had the highest incidence and the greatest association with harm. Nonetheless, further research is needed to understand and reduce infection control errors and consent/communication/confidentiality errors within the OR since these were frequently reported. If research focuses on the likelihood of harm, then documentation (including records, identification) should be added to this list of priorities. Outside of the OR, efforts could be directed at medication errors and patient accidents because these were frequently reported and resulted in harm when they did occur.

In conclusion, this report offers unique insights into the types of incidents occurring during cardiac surgical care in the UK. Although this study cannot provide data on incidence rates, it does reveal several event types that may pose the greatest risk to patients having cardiac surgery. Using the knowledge gained by this analysis, it will be important to develop systematic changes that may decrease the likelihood of future incidents and the ensuing patient harm. This knowledge must be shared with providers so they can learn and improve their performance. By focusing interventions on the events most likely to cause harm, clinicians may be able to reduce preventable harm in cardiac surgery, which is the ultimate goal of patient safety efforts. On the basis of the data we analyzed, the focus within the OR may be directed at medical devices/equipment, whereas outside the OR, it may be directed at medication errors and patient accidents. A follow-up study comparing our results with future results would be beneficial.


This work was supported by the Society of Cardiovascular Anesthesiologists Foundation (SCAF) as part of their Flawless Operative Cardiovascular Unified Systems Initiative (E.A.M., D.A.T., L.B., A.P.G., J.A.M. and P.J.P.); the National Health Service, National Patient Safety Agency in the United Kingdom (A.S. and J.C.P.); and the Agency for Healthcare Research and Quality (K08HS013904-02 to E.A.M., K01HS018762 to A.P.G.). The Flawless Operative Cardiovascular Unified Systems Initiative is a collaborative project of the Society of Cardiovascular Anesthesiologists, the SCAF, and the Johns Hopkins University Quality and Safety Research Group. The Flawless Operative Cardiovascular Unified Systems is funded exclusively by the SCAF. All of the work was completed at Johns Hopkins University.


The authors thank Christine G. Holzmueller, BLA(Hon), for her thoughtful review and editing of the manuscript.


View Abstract