Wrong patient, wrong procedure and wrong site surgeries are rare adverse events that are, in essence, systemic and have, in many instances, complex causes.1 However, such events are often severe, and the public view that they can be easily avoided, makes prevention a priority.23 At hospital level, it is essential to assess progress towards this objective and this requires combining several methods of evaluation.

First, the incidence of wrong patient, wrong procedure, and wrong site surgeries needs to be monitored. These outcome measures can be computed using denominator data that can be easily extracted from, for instance, computerised logbooks of operating theatres and numerator data from mandatory or voluntary incident reporting systems or from claims and complaints. However, both sets of data are likely to be unreliable, since under-reporting using these channels is common,4 and they are useless for detecting trends, given the low frequency of these events.5 Second, to overcome the first obstacle, periodic audits of compliance with verification protocols, such as those published by national patient safety organisations,6^–9 have to be carried out. Indeed, as exemplified by Shulman in a study on transfusion practice,10 they provide a convenient way of obtaining process measures that represent adequate surrogates to these outcomes. Audits of compliance can also help identify barriers to compliance with protocols. Finally, both outcome and process measures have to be compared with benchmarks. With regard to outcome measures, national monitoring systems provide numerator data only. For example, between January 1995 and December 2006, 532 wrong site surgeries were voluntarily reported to the Joint Commission on Accreditation of Healthcare Organizations sentinel events programme, representing the first category (13.1%) of all related adverse events.11 However, outcome measures are uncommon. Only two studies estimated the frequency of wrong site surgery and the rates ranged from one procedure in 28 00012 to one in 113 000.13 The frequency of occurrence of wrong patient and wrong procedure surgeries remains unknown. With regard to process measures, to our knowledge, there is no published estimate of compliance with prevention protocols for wrong patient, wrong procedure and wrong site surgeries.

Although not rigorously documented, wrong patient or wrong site surgery is sometimes triggered by wrong patient or wrong site anaesthesia.14 In our anaesthesiology service, instances of wrong patient or wrong site anaesthesia are reported through the institutional mandatory adverse event reporting system or through the departmental voluntary incident reporting system. Surgical procedures are documented in the computerised logbooks of operating theatres as well as in the anaesthesia information management system. From the beginning of 1998 to the end of 2005, no case of wrong patient anaesthesia occurred in a total of 181 710 surgical procedures (binomial upper 95% limit 1/60 643). In the same time period, 17 104 unilateral surgical procedures (such as a left-sided knee arthroscopy) were performed and four cases of wrong site anaesthesia (for instance, a femoral local anaesthetic block given on the wrong leg) were reported, corresponding to a rate of one in 4276 (binomial 95% CI 1/15 694 to 1/1670). None of these errors resulted in wrong site surgery (binomial upper 95% limit 1/5710). During the same period, however, it was not possible to obtain measures of compliance with any verification protocol. Indeed no such protocol was available either at national or at institutional/departmental level. In particular, there were no formal arrangements to prevent an anaesthetic being administered to the wrong patient or at the wrong site.

Therefore, we felt that prevention of wrong patient and wrong site surgery, not to mention accountability, demanded an intervention aimed at improving the way both patient identity and site of surgery checks were performed, while acquiring the ability to identify and correct deficiencies. This intervention took place in our programme for developing patient safety. It was based on two elements: (i) drawing up a verification protocol and (ii) conducting periodic audits to measure compliance while providing feedback.

CONTEXT OF THE PROBLEM AND STAFFING ARRANGEMENTS

Our anaesthesiology service operates in a 1200-bed university hospital. It is staffed by anaesthetists and nurse anaesthetists working in six distinct operating areas. According to our anaesthesia information-management system, anaesthetics are administered for around 23 000 procedures every year. Of these procedures, 1800 are carried out in the main outpatient operating theatre (ie, hand, orthopaedic, abdominal and plastic surgery) and 4400 in the main inpatient operating theatre (ie, cardiovascular, thoracic, orthopaedic, urological, and abdominal). The remaining procedures (eg, paediatric, obstetric and gynaecological) are performed in additional specialised operating suites.

At the time of our intervention, it was not mandatory for all patients to wear an identification wristband during the whole hospital stay, although a study had revealed that a vast majority of patients were in favour of such a requirement.15 However, two requisites to verifying patient identity and site of surgery had already been enacted at institutional level: all patients undergoing surgery had to wear an identification bracelet and the surgical site had to have the surgeon’s signature. The anaesthetic and surgical teams independently performed checks of patient identity and site of surgery.

ASSESSMENT OF PROBLEM

Development of the verification protocol

The verification protocol was developed for the anaesthesiology service by an interdisciplinary team consisting of doctors and nurses from anaesthesiology, surgical intensive care and orthopaedics. The protocol described how checks of patient identity and site of surgery had to be performed. The protocol also included suggestions on how to resolve discrepancies identified during the checks. The first requirement of the protocol was that the anaesthetist or the nurse anaesthetist in charge of a patient should perform checks on identity and site of surgery before administering the anaesthetic. If the patient was able to participate in the verification process, they were asked to provide their first and last names, their date of birth, and, when applicable, the site of surgery. To avoid suggesting an answer, questions had to be open. Following these checks, the identity data had to be compared with the information on the patient’s wristband and with the data provided in the operating theatre schedule and the patient’s medical record. The site of surgery check had to be compared with the surgeon’s check (signed by them) and with the information provided in the operating theatre schedule and the patient’s medical record. The protocol was distributed to all anaesthesiology staff members as a pocket-sized document.

Audits

Audits were conducted throughout the 9-month period from October 2003 to June 2004, and over 30-day periods in October 2004, March 2005 and October 2005. A nurse auditor carried out the audits in the morning, in both the main outpatient and inpatient operating theatres. The audits consisted of direct observations of the first contact between a patient and the anaesthetist or nurse anaesthetist, during which identity and site of surgery checks had to take place. About eight observations were made every day, and 11 criteria derived from the protocol were evaluated (table 1). In particular, the formal patient identity check first required staff to ask patients to state their identity, if they were able to participate in the verification process. It then required comparing the information obtained from patients with that from their wristbands, operating theatre schedule and patients’ records. A similar procedure was followed for the site of surgery, with information on wristbands being replaced with the surgeon’s check (signed by them). Data were entered into a standard data collection sheet. The date of the audit, patient’s participation in the verification process, and reasons for not performing a check were also recorded.

Statistical analysis

Compliance with the verification protocol was assessed over time as the percentage of observations that satisfied each audit criterion. 95% CIs were computed assuming binomial distribution. To study trends, we arbitrarily defined four 3-month audit periods. The initial 9-month period (October 2003 to June 2004) was split into three quarters (4th quarter 2003, 1st and 2nd quarter 2004), while the three subsequent 30-day periods (October 2004, March 2005, October 2005) together formed the fourth follow-up period. Pearson’s χ² test was used for overall comparisons between audit periods. Statistical analysis was done using Stata 8.2.

STRATEGY FOR CHANGE

The purpose of the audits was not withheld from the anaesthesiology staff members. After completing an observation, the nurse auditor was allowed, for educational purposes, to explain the protocol and discuss issues of inadequate checks with the professionals involved. Further feedback was provided by presenting the verification protocol and audit results at staff meetings, and in our institutional and departmental newsletters.

RESULTS

During the intervention, 1000 interactions between patients and anaesthetists or nurse anaesthetists were observed; in 985 interactions, patients participated in the verification process. Overall, compliance with all audit criteria significantly improved over time (p<0.001) except for “surgical site signed” (80.6%; CI 77.5% to 83.5%; 557/691). The improvement occurred mainly between the 4th quarter of 2003 and the 1st quarter of 2004 (table 1). During the follow-up period, over 90% compliance was reported for the two audit criteria: “Patient wearing wristband” and “Check of surgical site performed”.

The only reason reported for not performing a patient identity check was “I already know that patient” (the anaesthetist in charge of the patient had recently seen the patient during a preanaesthetic consultation) and represented 79.5% (CI 73.3% to 84.8%; 163/205) of instances. Only two excuses were provided when the site of surgery was not checked: “I already know that patient” (53.9%; CI 42.1% to 65.5%; 41/76) and “It’s the obvious surgical site” (21.1%; CI 12.5% to 31.9%; 16/76).

Between 1 January 2003 and 31 December 2005, a period during which we were conducting our intervention, 71 145 surgical procedures were performed, out of which 7688 were unilateral. There were no instances of wrong patient or wrong site anaesthesia.

LESSONS AND MESSAGES

During the intervention, which combined the dissemination of a verification protocol and periodic audits, compliance with all but one of the audit criteria noticeably improved and over 90% compliance was reached in two. This was primarily achieved by dealing with a procedural deficiency. It was customary for patients undergoing outpatient surgery to be directly admitted to the operating theatre without receiving a wristband, a practice amended after the first month of our intervention. Improvements were also obtained by modifying professional practice, as shown by the switch from closed to open questioning and the increased use of three different sources of information to cross-check a patient’s answers (patient’s wrist band/surgeon’s signature, operating theatre schedule and patient’s medical record).

Our study also revealed that to further improve compliance with checks of patient identity and site of surgery, two barriers have to be overcome:

• First, we must convince professionals that the often stated arguments for not performing a considerable proportion of patient identity and site of surgery checks (“I already know that patient”, “It’s the obvious surgical site”) are not valid from a patient safety perspective. Indeed, these arguments are contradicted by both potential and actual worst-case scenarios. For example, patients who are seen by an anaesthetist during the preanaesthetic consultation might be hospitalised on the same ward. As occasionally reported at our institution, an identity error can happen on the ward—before the patient goes to the operating theatre. Then, the anaesthetist in charge of the patient might not notice that the face he or she recognises and the name he or she remembers do not match. There are also actual examples of wrong site surgery where the wrong site is perceived as the “obviously” correct site by the professionals involved. For instance, two years before our intervention, a surgeon and an anaesthetist working in a large Swiss hospital were indicted in a well-publicised lawsuit about a patient in whom a leg with chronic ischaemia was removed instead of the one with acute ischaemia. The amputated leg presented extensive skin lesions whereas the opposite one was apparently healthy.16 Interestingly, even repeated presentations of these scenarios during our intervention period were unsuccessful in convincing professionals that patient identity and site of surgery checks should not be bypassed under any circumstances. Communication problems between patients and professionals could explain why such rational arguments do not work. Indeed, although patients are today encouraged to speak up in order to prevent errors,17 very little is being done to train professionals in presenting safety issues to patients. Consequently, some professionals might feel that the abrupt questions required to satisfy safety protocols such as “Can you tell me your name?” or “What is the site of your surgery?” could be perceived by patients as a lack of professionalism (professionals should know their patients) and rather daunting. Noticeably, we observed that patients were more often asked for the site of their surgery than for their identity. Anaesthetists and nurse anaesthetists could perceive questions related to surgery as being more acceptable to patients if asked by anaesthesiology professionals.

• Second, we have to improve our collaboration with the surgical services. Although orthopaedic surgeons had participated in the development of the protocol, we failed to get all surgical services actively involved in its implementation. This could in part explain why the proportion of missing surgical site marks remained stable during our intervention. But, as revealed by an observational study on the human factors that limit the implementation of verification protocols, it is likely that the low proportion of patients having their surgical site marked is in essence a systemic issue: in many instances, such protocols might not “account for the dynamic complex operating room environment”.18 The lack of collaboration with surgeons also hindered the implementation of some of the recommendations made by various agencies around the world.2689 This was especially true for the team “time out”, a final verification that takes place immediately before starting the procedure and that has to be collectively performed by the anaesthesiology and surgical staff. In these circumstances, a national or institutional regulation could contribute to the development of a more concerted approach.

Our study had three main limitations. It did not include a control group and could therefore have been subject to undetected bias. However, during the intervention period, we could not identify any factors that could have possibly influenced professional behaviour (for instance, recently published national or institutional policies; new and publicly discussed cases of wrong patient or wrong site anaesthesia). Direct observations could have altered professional behaviour, since the purpose of the study was overt. Consequently, our compliance results could be overestimated. However, the potential impact of this observational bias should not be overstated, since strictly conforming to a detailed protocol in presence of an auditor while getting away with it otherwise seems, in the long run, unrealistic. This would require professionals to continue to remember the protocol perfectly step by step with diminishing practice as the frequency of audits and feedback decreases in the follow-up period of our intervention. The opposite approach—attempting to avoid this bias by resorting to disguised observations—could have limited the impact of our intervention since audit and feedback are an effective way of changing professional behaviours, especially when initial compliance is low.19 Finally, however encouraging our results, one should be able to assess whether the level of safety achieved in the prevention of wrong patient and wrong site anaesthesia is satisfactory. Concerning outcome measures, no case of wrong patient or wrong site anaesthesia was observed during our intervention. However, because of uncertainty limits, this did not indicate that everything was all right, as Hanley and Lippman-Hand point out.20 With regard to process measures, compliance data that could be used as a benchmark are almost non-existent. So far, only studies relating to identification wristbands are available in the literature. These studies suggest that large variations exist between countries and institutions. For instance, one survey carried out over a 2-year period in 217 institutions across the USA revealed that on average, 98.2% patients were wearing a wristband.21 But in another study conducted in a NHS trust, this proportion was initially at 65%. It increased to 98.5% in less than a year following a dedicated intervention.22 Similarly during our intervention, the proportion of patients wearing a wristband quickly rose from 81.8% to 99.6%.

Comparing compliance between institutions also raises the issue of the definition of audit criteria. In our study for example, formally performing a check required that staff ask patients able to participate in the verification process to confirm their identity and, when applicable, the surgical site. Consequently when one of these two questions was omitted, the corresponding check was deemed not performed, even if some comparisons, for instance between mark of surgical site and operating schedule, were made. Such a stringent criterion could explain why observed compliance was low especially at the beginning of our intervention when professionals were not familiar with the verification protocol.

However, ultimately full compliance with compulsory checks of patient identity and site of surgery is a necessary condition for improved patient safety, but not a sufficient one. Indeed, full compliance with a verification protocol only allows one to achieve its intrinsic reliability; that is the proportion of errors the protocol is able to capture when rigorously performed. To our knowledge, this proportion has been evaluated in only one study, which has suggested that rigorous application of the JCAHO protocol7 would have prevented no more than two-thirds of wrong site surgeries.13 So far nothing is known about the influence on intrinsic reliability of the many different features1315 that make up verification protocols developed by institutions and services. Consequently, most verification protocols are based on common sense rather than on scientific evidence—an approach to patient safety that is strongly criticised.23 This observation should encourage the design and validation of verification protocols having high intrinsic reliability using, for instance, barcode or radiofrequency identification. Meanwhile, measuring compliance with written verification protocols and providing feedback seem the only practical ways of effectively preventing wrong patient and wrong site surgery, improving patient safety, and preparing for the next step. Indeed, good manual systems have to be in place before information technologies can be implemented. Sustained periodic audits and feedback could contribute to this objective through the identification of barriers hindering compliance with protocols.