Safer Surgery 324 (Figure 19.1). Other relevant data were recorded, including operative duration (rst incision to nal closing suture), tourniquet time and the composition of the surgical team. For the purposes of the study, risk was classied at levels; low risk for primary TKR procedures,and high risk for TKR revisions. One operation was a TKR revision, but was classed at low risk as it involved only the removal of the existing prosthesis. Another operation was a primary knee replacement, but was classed as high risk as it required instruments, prostheses and techniques used in revision operations. Events were selected for analysis if they were judged to have increased the duration or difculty of the operation, increased the risk to the patient, or increased the demand for resources. They were all categorized individually as minor failures. A major failure category was reserved for events which were approaching an incident or accident (see Box 19.1). Where major failures or unusual or complex minor failures occurred, brief reviews were conducted with relevant theatre team members at a convenient time following the operation to ensure that the supporting specialist information had been recorded. Video evidence was used to check the results of the observers. Minor failures were grouped into 20 types previously dened for paediatric cardiac surgery (Catchpole et al. 2006) (Table 19.2). Figure 19.1 Video equipment conguration for orthopaedic surgery Source: Catchpole et al. (2005) Observing Failures in Successful Orthopaedic Surgery 325 Failure Description and example Absence Lack of personnel when required. Example: circulating nurse is absent when scrub nurse needs more suture material. Coordination/ communication failure Failures in task coordination and communication between individuals. Example: surgeon asks for the drill, but the scrub nurse is busy doing something else. Decision-related sur gical error Surgeon fails to make the appropriate decision. Example: the surgeon nds the tibial cut is sloping, after the assistant surgeon has repeatedly expressed his concern. Distraction Disturbance from external sources not related to current case during a critical period. Examples: (i) telephone rings in theatre; (ii) another nurse enters theatre and distracts the scrub nurse. E quipment/ workspace management failure Failures in the organization of workspace and equipment. Examples: (i) the sur geon tries to use the saw, but it is not plugged in; (ii) x-rays are displayed the wrong way round. Equipment conguration failure Failure to use or operate equipment appropriately . Example: (i) intra-medullary rod not inserted far enough into femur; (ii) cutting block moves as pins are hammered in. Equipment failure Inter -operative equipment failure. Examples: (i) sutures break; (ii) tibial tray is bent when it comes to being used. Expertise/skill failure Failures associated with lack of expertise or skill in trainees. Example: (i) assistant surgeon does not know how to use the bone saw correctly; (ii) not enough cement is applied to the prosthetic. E xternal resource failure Failures in elements of the external organization to provide equipment or human resources. Examples: (i) piece of equipment is missing from the standard set; (ii) correct pins unavailable from prosthetic manufacturer . Patient-sourced procedural difculties Features of the patient that make the planned procedure more challenging to carry out than would be expected from the pre- operative diagnosis. Example: (i) patient apnoeic, requiring re-intubation (ii) not enough femur left to make box cuts. Planning failure Failure to anticipate or discuss future task requirements. E xample: surgeon consults patient notes after the start of the operation. Table 19.2 Descriptions and examples of minor failure types Safer Surgery 326 Minor Failures The two observers found 327 minor failures in the study, varying between 10 and 50 in each operation, with a mean of 23.3 per operation (95 percent C.I. ± 7.82). The number of minor failures in an operation showed a moderate relationship with duration (rho(14)=0.678, p<0.01) and tourniquet time (rho(14)=0.788, p<0.005). Failure Description and example Pre-operative diagnosis failure Failure to provide accurate diagnosis prior to operation. Example: sur gical team have decided upon an implant from one manufacturer, but the x-rays show the previous implant to be from another manufacturer. Procedure-related error Procedural errors by surgeon, assistant surgeon, anaesthetist, scrub nurse, or circulating nurse. Examples: (i) surgeon forgets to plug the intra-medullary hole (ii) cement mixing time is not recorded. Psychomotor error (general) H andling errors. Example: sur geon drops prosthesis while applying cement. Psychomotor-related surgical error Technical manipulation errors by surgeon. Examples: (i) assistant surgeon gets forceps tangled in sutures; (ii) assistant surgeon hits nger of surgeon with hammer. R esource management Failures in the organization of available people or things in the operating theatre. E xample: surgeon leaves assistant surgeon to close without conrming ability to do so. Safety consciousness Failures to observe basic elements of patient safety. Examples: (i) mask is not tted on entry to theatre; (ii) sur geon does not have eye protection while making bone cuts. Team conict Team members have differing opinions, or give conicting commands, that are not resolved. E xample: scrub nurse and assistant surgeon ar gue over procedural requirements. Unintended effects on patient Unplanned problems arising with the patient as a result of the treatment. E xample: blood pressure increases unexpectedly after bag of uids is changed. Vigilance/awareness failure Failures to notice immediately important aspects of the task or the patient’ s condition. E xample: anaesthetist does not notice a drop in blood pressure. Table 19.2 Concluded Observing Failures in Successful Orthopaedic Surgery 327 There was also an effect of operative risk on the number of failures encountered in each operation, with an independent samples t-test assuming unequal variances showing signicantly more minor failures in high risk operations (t(3.6)=-6.53, p<0.005), though the small number of TKR revisions limits the condence of this relationship. One operation was considered to have contained a major failure sequence where the tibial cut was misaligned (see Box 19.1). Though this was identied and resolved by making further tibial cuts, it was considered a major failure because Box 19.1 Description of a major failure A 66 year old male was anaesthetised and prepared for a primary total knee replacement operation. The surgeon was already aware that the knee was in an advanced state of wear, and might require additional prosthetic components usually required only for second stage TK R revisions, so the corresponding instrument trays and prostheses were kept on standby. The sales representative for the prosthetic manufacturer was present for the operation, as was normal for TKR revisions, to assist in the selection and conguration of the implants. On this occasion, the surgeon also chose to use a new computer navigation device that was on trial from the prosthetic manufacturer. This was normally used for standard primary TKR operations, and had not previously been used by the team for this type of operation. Though the surgical team were all experienced and familiar with one another, early in the operation the scrub nurse showed signs of uncertainty both about the use of the computer equipment, and the surgical direction being taken. The sales representative offered his support to the scrub nurse, but his advice conicted with instructions from the surgeon, and the scrub nurse became confused about what was required next. At the same time, the surgeon was struggling to operate the computer navigation system, becoming increasingly frustrated with it. The sales rep also tried to support the surgeon, but was unable to support both scrub nurse and surgeon at once. He was also trying to coach a second sales representative at the same time. There was a high level of conversational noise in the operating theatre. As the operation progressed, the scrub nurse began to unpack several extra trays of instruments required for the additional implants, with help from the sales rep. While the rep was outside the theatre, the surgeon removed one of the navigation sensors, and on his return the rep expressed concern that this may have upset the calibration of the navigation system. A few minutes later, the assistant surgeon became concerned about the tibial cut that had been made. The surgeon and sales rep dismissed his concerns and instead teased him for being overly anxious. In the following minutes there were a number of other confused exchanges between the sales rep, the scrub nurse and the surgeon. Sixty three minutes after the start operation the surgeon noticed that the tibial plateau had not been cut properly and was on a slope. The operating theatre went silent. The assistant surgeon then conrmed that this was his earlier concern, and appropriate cuts are made to rectify the problem. Following a number of further co-ordination difculties, the operation was eventually completed after 126 minutes. Safer Surgery 328 it demonstrated a substantial deviation from the normal operative course which had a direct affect on the patient as it required the removal of more bone than was intended. This event was indirectly precipitated by a number of other issues, including uncertainty, high workload, task requirements, non-technical errors and the introduction of new technology into the operating theatre. This event occurred in the higher risk operation, and also had the greatest number of minor failures in any operation observed. The 327 failures were classed by 20 failure types, and the average number observed per operation is shown in Figure 19.2. Approximately half the minor failures observed were distributed among three failure types, reecting either aspects of theatre culture and environment, or the more difcult elements of the equipment and procedure. Distractions were the most numerous failure, accounting for nearly twice the number of the next most frequently observed failure. The greatest distracters were communication devices, including 16 wrong numbers or calls for people not in the theatre, 26 other non-case-related telephone calls (case-related calls were excluded from the data), 10 mobile phone calls and 9 Figure 19.2 Mean number of minor failures per operation by type Observing Failures in Successful Orthopaedic Surgery 329 instances of bleepers/pagers sounding in theatre. The demands for intra-operative teaching, as well as the presence of newspapers and other non-case-related reading material during the operation, also contributed to this high level of distraction. Several instances were observed where these distractions coincided with parts of the procedure particularly susceptible to interference, such as equipment counts and timing of bone cement curing. The demands to organize, prepare and exchange different pieces of equipment (jigs, drill bits, saw blades, cutting blocks, implants) during the operation is reected in the high number of equipment and workspace management failures where equipment was needed and was not immediately available (such as nding the drill not plugged into the air line when attempting to use it), or where it was placed in a physical conguration that compromised the operation of the equipment (such as tangled diathermy wires). There were also more than 40 instances suggesting reduced safety consciousness, mostly relating to mask discipline – failures to wear masks appropriately at all times – in theatre, but also including one event where the scrub nurse left the theatre and returned while still scrubbed. Coordination and communication problems were frequent, suggesting problems either in the distribution of important information within the surgical team, or to the timely execution of procedural sequences requiring interaction between individuals. Procedure-related errors, expertise/skill failures and equipment conguration failures also reect the close-coupling of procedures, equipment and expertise in this type of surgery. In particular, less experienced surgeons and scrub nurses suffered expertise or skill failures because of tightly coupled tasks, especially in the more demanding TKR revision surgery. Failures in equipment, vigilance/ awareness failures and absences that had an impact on the operation were not as frequent, but still occurred in over half the operations. However, technical skill- based issues including decision-related error, diagnosis failures and psychomotor errors were considerably less frequent, along with more undesirable elements of the operation, notably unintended effects on the patient, and team conicts. Applying the Failure Source Model As failures sometimes reect individual errors, failures in group processes, threats outside theatre that affected events in theatre, or combinations of all three, the failure source model (Figure 19.3) was applied. This allowed the description of events in terms of the number and types of failures that could be observed, and the systemic threats and human errors that they represented. Equipment, workspace and resource threats and organizational/cultural threats were associated with four different minor failure types each. Task threats were associated with eight different minor failure types and patient threats were associated with three minor failure types. Technical errors were associated with ve different minor failure types and non-technical errors were associated with nine different types of minor failure. As it was not always possible to observe the specic source of the failure in Safer Surgery 330 every case, the model assumed that all potential sources had contributed in equal proportions. Applying this failure source model allowed the generation of a threat and error prole for each operation (Vincent et al. 2004) which was used in the nal systems analyses. From the failure source model it was possible to estimate the rates of systemic threats and human errors. Figure 19.4, top panel, shows the mean failure rates in terms of threat type within each operation divided by the length of the operations. Cultural and organizational threats were the most frequently encountered, and are more frequent in higher risk operations. Task threats and equipment threats occur at approximately the same rate, regardless of risk, with patient threats the least frequent. This general pattern would be expected given the process and equipment driven nature of the TKR surgery where patients are relatively invariant. Figure 19.4, bottom panel, shows the minor failure data treated in a similar way to estimate human error rates. As might be expected, error rates were higher for high risk operations. Also, non-technical errors, occurring once every 10–15 mins, were Figure 19.3 Failure source model which links observable minor failures (small boxes) and common systemic causes (large boxes) Observing Failures in Successful Orthopaedic Surgery 331 considerably more frequent than technical errors, which occurred only once every 30–50 mins, though the latter appear to be less variable between operations. Comparison Between Dual Observers All operations were dual-observed to examine the agreement between observers with two very different backgrounds. Observer 1 was a human factors practitioner, experienced in observational methods in the operating theatre, human performance measurement and the evaluation of non-technical skills. Observer 2 was a research scientist with prior experience of working in operating theatres and with some prior knowledge of non-technical skills, but no specic human factors or observer training. The 17 time markers anchored to specic points in the primary TKR Figure 19.4 Mean rates of threats (top panel) and errors (bottom panel), with 95 percent condence intervals Safer Surgery 332 operations allowed an assessment of the level of understanding that the observers had of the basic stages of the operation. Correlation between time markers in the ten dual-observed primary TKR operations was almost perfect (rho(151)=0.998, p<0.001). Although all 17 times were not always recorded in each operation, this suggests that both observers were in close agreement about when each stage of the operation occurred. Of the 327 different failures recorded in the dual-observer operations, 72 were recorded by both observers, 193 were recorded by observer one only, and 62 were recorded by observer two only. Overlap between observations was often low, accounting for between 0 percent and 42 percent of the total number of failures observed in each operation. A Bland-Altman plot (Bland and Altman 1986) was used to examine inter- observer agreement, and can be found in Figure 19.5. Observer one generally saw the same or twice as many failures as Observer two. There were three outliers where the number of failures noted by Observer one was considerably more than the number noted by Observer two. One was the computer-navigated operation, which was a technique unfamiliar to Observer two. The other two outliers show a oor effect, where the number of intra-operative failures was generally small. One outlier was found in the operation with the lowest number of failures, where Observer two did not see any failures independently, while in the other outlying operation there was no overlap between observations, suggesting that both observers had a different focus of attention. This might be expected in operations where the number of intra-operative failures was small. Though both observers Figure 19.5 Bland-Altman plot for agreement between two observers Observing Failures in Successful Orthopaedic Surgery 333 viewed events differently, and both under-reported the total number of observable intra-operative failures, both demonstrated a good understanding of the process and the observations from both remained similarly proportional across the range of operations viewed. Thus, though different in absolute terms, judgement between operations was consistent between observers. While each individual observer did not observe all intra-operative failures, there was general consensus between the observers about the success of each operation. Both were able to identify a considerable number of failures, with the small overlap between observations suggesting differences in experience and focus of attention between observers. Intra-observer reliability might be improved by the explicit denition of salient events, and the present study could undoubtedly inform a comprehensive approach to the structured observation of pre-dened intra-operative failures. However, given the unpredictable nature of errors and the importance of understanding the context in which they happen, explicit tools are unlikely to be sufcient. Rather, we would strongly advocate the value of implicit error capture conducted by experienced observers, which should then utilize a structured assessment protocol. Since the human factors practitioner with experience in surgery was able to observe nearly twice as many failures as the observer with surgical experience but little human factors training, it is also clear that observation by non-medically trained observers is vital for identifying and reducing medical errors. Understanding the threats and errors encountered by surgical teams in this way makes possible the diagnosis of these important but often neglected events. Discussion Intra-operative events that resulted in small increases in the duration or difculty of the operation, risks to the patient, or demands for resources were frequent, and varied considerably between operations. Distractions, equipment management problems, safety consciousness and coordination and communication problems were the most frequently occurring types of failure. Decision-making and diagnostic errors were the least frequent type of failure, reecting the skill and competence with which the individuals studied carried out the technical demands of their tasks (or possibly a failure of the observers to detect failures in these areas). The variability in the number of failures, given that the team varied so little, suggests that these failures must have their source in variable systemic conditions. Higher risk operations were more likely to include a greater number of errors than lower risk operations, and though sample size limits the condence of this result, it conrms previous ndings in paediatric cardiac surgery (Catchpole et al. 2006). Several team members suggested that operations were more frequently problematic with other surgeons. Consequently, we believe the data presented here are likely to provide a conservative estimate of the likely incidence of intra- operative failures in orthopaedic surgery across the UK. . cardiac surgery (Catchpole et al. 2006) (Table 19.2). Figure 19.1 Video equipment conguration for orthopaedic surgery Source: Catchpole et al. (2005) Observing Failures in Successful Orthopaedic Surgery 325 Failure. of surgery. In particular, less experienced surgeons and scrub nurses suffered expertise or skill failures because of tightly coupled tasks, especially in the more demanding TKR revision surgery. . after the start of the operation. Table 19.2 Descriptions and examples of minor failure types Safer Surgery 326 Minor Failures The two observers found 327 minor failures in the study, varying between