INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: bronchoscopy; simulation; training; education

Procedural skills are receiving increased attention because of their importance to patient care and because more rigorous standards are being proposed for the issuing of credentials (1- 3). However, surveys of pulmonology fellows and attending physicians demonstrate significant variation in bronchoscopy experience during training and subsequently during practice (4). In a survey of pulmonary fellows, 28% had performed fewer than 50 bronchoscopies per year (4). Lack of adequate training can lead to underutilization of effective techniques, such as transbronchial needle aspiration (TBNA) (6). In addition, the number of bronchoscopic examinations being done in the United States is decreasing and this may have an adverse impact on future pulmonary fellowship training (7).

With the decreasing number of bronchoscopies being performed and the ongoing development of new technologies and procedures, there is an increased need to develop better ways to teach bronchoscopy and assess bronchoscopy skill (5). Previous reports on pulmonary fellows' impressions of bronchoscopy training suggested that simulation might be useful in addressing some of these issues (4, 8, 9). We therefore sought to further develop and validate a bronchoscopy simulator.

To effectively evaluate and validate any type of simulation requires that at least two requirements be met (9). First, there must be effective skill transfer from real-life to the simulation. How well skills acquired by performing actual bronchoscopy transfer to the simulation is important if simulators are to be used to measure the bronchoscopy skill of trainees. Presumably, expert bronchoscopists should outperform novices on the simulator. This is also important because skill transfer from actual to simulated environments is an effective measure of how well the simulation mimics the actual activity (10). The second requirement is that there must be effective skill transfer from the simulation to actual bronchoscopy, such that practice with the simulator results in improved performance. Transfer of skill from simulation to reality is not always equal to transfer of skill from reality to the simulation, so each of these must be studied separately (10).

To assess a bronchoscopy simulator, we performed an observational study to first test whether the bronchoscopy simulator would be able to distinguish between novices and experts on the basis of various measures of bronchoscopy performance. We then used the bronchoscopy simulator in a small randomized-controlled trial to determine whether simulation training for new bronchoscopy fellows would prove superior to conventional bronchoscopy training.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Description of Simulator

The AccuTouch Flexible Bronchoscopy Simulator (Immersion Medical, Gaithersburg, MD), consists of a proxy flexible bronchoscope (FB), a robotic interface device, a computer with monitor, and simulation software. These components combine to create a realistic and immersive training environment for learning and practicing flexible bronchoscopy. The simulator is now commercially available.

As shown in Figure 1, the user inserts the proxy bronchoscope into the robotic interface device. The bronchoscope is supposed to mimic the feel of an actual bronchoscope. The interface device tracks the motions of the FB and reproduces the forces felt during an actual bronchoscopic procedure. The proximal end of the interface is shaped like a human face with a port to insert the FB through one of the nasal passages. The FB tracks the manipulations of the tip control lever, the suction button, and video buttons. In addition, instruments are tracked as they are manipulated in the working channel. This allows for biopsies and other diagnostic and therapeutic procedures to be performed on the simulator. For the purposes of this study, no biopsies or diagnostic procedures were performed.

View larger version (150K): [in this window] [in a new window]   Figure 1.   The flexible bronchoscopy simulator. A: Proxy FB; B: mannequin face with port for scope insertion; C: interface device.

The monitor displays computer-generated images of the airway as the user navigates through the virtual anatomy. Figure 2 demonstrates an example of the virtual anatomy at the level of the carina. The three-dimensional computer-generated model of the airway was constructed using the National Library of Medicine's Visible Human male data set (12). This data set consists of digital axial anatomic images at 1.0-mm intervals with associated computed tomographic (CT) and magnetic resonance imaging (MRI) images. Texture maps based on videotapes of actual bronchoscopic images were added to give the mucosa of the virtual airway a realistic look. Adjusting the virtual airway to reflect common anatomic variations generated a variety of patient cases.

View larger version (72K): [in this window] [in a new window]   Figure 2.   Bronchoscopic view of simulated airway.

In addition to being anatomically correct, the virtual patient also behaves in a physiologically realistic manner. The patient breathes, coughs, bleeds, and exhibits changes in vital signs. Complications such as lidocaine toxicity causing the patient to seize or develop a cardiac arrhythmia are also programmed in.

The simulation software records all the actions of the user and stores this information in a database. Information that is collected and displayed includes time of procedure, time that the bronchoscope is in "red-out," number of times that the bronchoscope tip collides with airway walls, the percentage of bronchial segments entered, and the amount of lidocaine used. Red-out occurs when the FB tip presses against the mucosal wall. The light reflected from the mucosa has a red color that prevents adequate visualization of the airway. Appropriate use of lidocaine reduces cough and thus wall collisions. Excess use of lidocaine will cause toxicity. Similarly, proper use of suction allows for a thorough examination of the airway without wall collisions or red-out.

The software runs on a personal computer with a 500 MHz Intel Pentium III Processor and Microsoft Windows NT. The computer also requires a graphics accelerator card.

Observational Trial

The objective of the observational trial was to determine whether a bronchoscopy simulator would be able to distinguish between novices and experts on the basis of various measures of bronchoscopy performance as assessed by the simulation program. This portion of the study was conducted in three university or university-affiliated hospitals, each with its own pulmonary fellowship program.

Study groups and sample size.Expert bronchoscopists were defined as pulmonary attending physicians who had performed more than 500 flexible bronchoscopic examinations. Intermediate bronchoscopists were defined as pulmonary or pulmonary-critical care fellows who had performed more than 25 but less than 500 flexible bronchoscopic examinations. Novice bronchoscopists were defined as being familiar with the anatomy of the tracheobronchial tree, but who had never performed a bronchoscopy. This group included nonpulmonary fellows, bronchoscopy nurses, and second-year and third-year medical residents.

Experience levels were divided in this way based on consideration of within-group uniformity and sample size limitations. The "expert" bronchoscopists at the study sites all had performed more than 500 bronchoscopies. We also ensured that novices truly had no prior exposure to bronchoscopy. This resulted in a fairly wide "intermediate" category of 25 to 500 procedures. We chose a minimum of 25 bronchoscopies in this category to ensure that all "intermediate" fellows would have at least enough experience to be fairly adept at bronchoscopy, given that there is significant interindividual variation in bronchoscopy aptitude. Additional division of the intermediate category into smaller subgroups was not possible because of the small total sample size available. The total sample size available for the intermediate group consisted of all second- and third-year fellows. Dividing the intermediate group into two smaller groups would have resulted in two subsets, each too small in size to make meaningful comparisons.

Study coordinator.Each study site had a study coordinator present to instruct participants, ensure compliance with the protocol, and to collect data. The study coordinator was not a participant of the study.

Study protocol.The study participant was given instructions by the coordinator using a standardized protocol on the goals of the bronchoscopic examination and on the use of the simulator. Participants were told that total procedure time, percentage of bronchial segments entered, number of collisions, amount of time in red-out, suctioning, and amount of lidocaine used would be measured. The participant was given an illustration of the tracheobronchial tree to review the segmental anatomy. The participant was given up to 5 min to review the anatomy and had free access to an illustration of the tracheobronchial tree during the protocol to ensure more uniform knowledge between pulmonologists and nonpulmonary physicians.

The study participant was allowed 5 min to become familiar with the simulator by performing a bronchoscopic examination on a sample case. Data during this "play" time with the simulator was not collected. The purpose was to acquaint the participant with the simulator's operation. The participant then performed two different cases in the same sequence. Subjects were told to ignore any pathology that was found during the examianton with a maximum time limit of 20 min per bronchoscopic examination. The participant was instructed to visualize each segment of the bronchial tree and no feedback was given during the procedure. The procedure was ended when the subjects felt they had visualized the entire bronchial tree or at the 20-min time limit.

Demographics collected.Demographic data collected included medical specialty, level of training, previous experience performing or observing flexible bronchoscopy, exposure to bronchoscopy simulators, age, sex, computer experience, and level of video game experience. Computer experience was self-rated as either very experienced, somewhat experienced, inexperienced, or never used a computer. Level of video game play was self-rated as often (few times per week), occasional (few times per month), seldom (few times per year), or never.

Outcomes.The simulator measured total procedure time. This was measured from the moment of insertion of the FB into the "nose" of the robotic mannequin to the moment of removal of the scope from the "nose." The simulator also measured percent of bronchial segments visualized. A bronchial segment was "visualized" during the procedure if the "bronchoscope" passed beyond a certain point in the bronchial tree. Wall collisions, total amount of lidocaine used, and suctioning were also recorded. Finally, the simulator measured amount of time in red-out.

Statistical methods.A frequency table was generated for the values of the class variable levels of bronchoscopy experience (novice, intermediate, or expert), number of bronchoscopies performed, sex, and level of video game play. The number of bronchoscopies performed was dropped from further analysis because its frequency showed groupings that paralleled the level of bronchoscopy experience. Because time in red-out had a skewed distribution, the log₁₀ transformation was used in the analysis to achieve a normal distribution with equal standard deviations, which is required for analysis of variance (ANOVA).

A separate comparison of the outcome variables total procedure time, red-out time, suction time, total lidocaine, percentage of segments entered, and total collisions was done across the class variables level of bronchoscopy experience, sex, and level of video game play using three-way ANOVA. One-way ANOVA was carried out using only those class variables that were related to outcome variables. The Student-Newman-Keuls multiple comparison test was used to determine which bronchoscopy experience levels differed from one another.

Results were analyzed first with each performance on the simulator by each participant as a separate record, and then using the average performance between Cases 1 and 2 for each participant. There was no quantitative difference in outcomes between these two methods. Intraclass correlation coefficients for the outcomes of total time, time in red-out, suctioning, lidocaine use, percentage of segments entered, and collisions were done to compare the differences in performance between Case 1 and Case 2 for individual participants. There was no significant difference between cases. Subsequent analysis was therefore reported for Case 1 and Case 2 combined.

Randomized-controlled Trial

After validation of the simulator's ability to measure real-life expertise and identification of what outcome variables were significant, we performed a randomized-controlled trial to determine whether training new pulmonary fellows on the bronchoscopy simulator would lead to earlier attainment of expertise as compared with traditional training methods. The outcome measure was performance quality of pulmonary fellows during their first two real-life "solo-bronchoscopies."

Study groups and sample size.The limitations of yearly fellow recruitment limited enrollment in our randomized-controlled trial. This limited our sample size and thus limited the statistical power of the analysis. Six first-year pulmonary fellows were randomized to two arms, one using the simulator (SIM group) and one with traditional training (control). All pulmonary fellows had no previous exposure to the bronchoscopy simulator. No first-year pulmonary fellows had performed any previous bronchoscopies. Both groups underwent a core curriculum for bronchoscopy training. The SIM group then received additional training upon the bronchoscopy simulator, whereas the control group received additional conventional training.

Core curriculum.The core curriculum consisted of a lecture series reviewing indications, contraindications, risks, complications, and methods of performing bronchoscopy. This was followed by lectures on conscious sedation. A video of real-life bronchoscopy was then observed. Written materials included a bronchoscopy textbook with initial chapters for review selected. All lectures and videos were given with both SIM group and control group present at the same time to avoid bias. All fellows simultaneously received orientation to the bronchoscopy laboratory itself, with a hands-on chance to examine the bronchoscope controls.

Control group training.Upon completing the core curriculum, the control group observed and assisted with bronchoscopy at a county hospital for 1 mo. Teaching in the bronchoscopy laboratory consisted of first observing bronchoscopy, followed by being allowed to remove the bronchoscope, and then to pilot the bronchoscope without performing biopsies. This was the standard teaching arrangement that had been used for many years. The three pulmonary fellows in the control group were exposed to 10 to 15 bronchoscopies each during a 4-wk period. Upon completion of this 1-mo introduction, the control group performed their first solo-bronchoscopy as described subsequently.

Simulator group training.Upon completion of the video, the SIM group performed 20 simulated bronchoscopies on the AccuTouch Flexible Bronchoscopy Simulator. Each simulation bronchoscopy was performed with a supervisor to operate the computer but without instruction or aid from the supervisor. Fellows in the SIM group were told that the objective was to visualize each segment of each lobe to the subsegmental level (i.e., enter into the right middle lobe [RML], lateral segment) and to do this in as fast a manner as possible while still being complete. Participants were told that total procedure time, percentage of bronchial segments entered, number of collisions, amount of time in red-out, suctioning, and amount of lidocaine used would be measured. Upon completely visualizing the endobronchial tree, they were to remove the bronchoscope. The fellow determined the point at which the airway was "completely visualized"; they were given no clues as to whether or not they had fully evaluated the airway during the procedure. If they could not finish the simulation within 20 min, they were told to discontinue at the 20-min time limit for evaluation. Upon completion of each simulation performance, a computer evaluation was shown to the fellow. This included the total procedure time as well as a list of segments entered and missed. No recommendations were made on how to improve by any faculty or other staff member. This was done in order to simulate the fellow training without faculty supervision.

Each fellow performed the 20 simulated bronchoscopies in three to four sessions. Each session consisted of at least five simulated cases per session to a maximum of 10. This was done to simulate a fellow training on the simulator for 2 h per day for 4 d. Five different cases were available for simulation. Cases were done in order from 1 to 5 and then repeated. Upon completion of this simulated introduction, the control group performed their first solo-bronchoscopy as described next. They received no other orientation on "real-life" patients before their solo-bronchoscopy assessment.

The simulator measured total procedure time, percentage of bronchial segments visualized, wall collisions, total amount of lidocaine used, suctioning, and time in red-out.

Performance quality during solo-bronchoscopy.Performance quality was measured for both groups during their first two solo-bronchoscopies on real patients. Each fellow was told before the solo-bronchoscopy that they would be evaluated in terms of accuracy, speed, and subjective quality of the bronchoscopy. Fellows were told that the objective was to visualize each segment of each lobe to the subsegmental level (i.e., enter into the RML, lateral segment) and to do this in as fast a manner as possible while still being complete. All fellows were told to do the right side first, and then proceed to the left. Participants were told that total procedure time, percentage of bronchial segments entered and correctly identified, and subjective performance would be measured. Upon completely visualizing the endobronchial tree, they were to remove the bronchoscope to the midtrachea and tell the attending that they had finished. Fellows were told that if they had not finished their bronchoscopy by 20 min or if the patient had any difficulty, the evaluation would be ended. The fellow was to determine the point at which the airway was "completely visualized"; fellows were given no clues as to whether or not they had fully evaluated the airway during the procedure.

Fellows were told to identify each airway visualized as they performed the bronchoscopy by using the position of the airway on the video monitor relative to a clock-face. If there was doubt as to what the fellow was identifying, the attending physician pointed to it with a pencil and the fellow identified it. Fellows were allowed to change their answers as frequently as they liked, with their last answer being counted for that particular segment. Fellows were not told whether their answers were right or wrong during the bronchoscopy.

The fellow, under the attending physician's supervision but without prompting, did all local anesthesia and conscious sedation. All fellows were told to use a combination of meperidine and midazolam for conscious sedation. The fellow chose the amount and timing of medications. All procedures were performed with an Olympus P240 bronchoscope. To eliminate variation caused by nasal anatomy, the attending physician passed the bronchoscope through the nose and vocal cords, to position the bronchoscope in the midtrachea with the anterior of the patient being at 12 noon on the video bronchoscope. The fellow was then given the bronchoscope and told to begin. This was considered as time zero. During the procedure, no advice, prompting or guidance was given to the fellows to facilitate their performance.

Outcome measures.Quantitative outcome measures included total procedure time in seconds, the number of segments visualized, the number of segments visualized and correctly identified, number of coughing episodes, amount of lidocaine used for topical anesthesia by the fellow, total amount of midazolam used, and total amount of meperidine used. The quantitative bronchoscopy quality score was defined as (% of segments visualized and correctly identified)/(total time in seconds).

A qualitative overall bronchoscopy score, using a 1 to 10 scale with 5 being average and 10 best, was given to each performance by the bronchoscopy nurse who was blinded as to the type of training received by the fellow. The nurse was told to base the score on the overall quality of the bronchoscopy in terms of patient preparation and anesthesia, patient comfort, speed, skill, coughing, and teamwork and communication with the bronchoscopy nursing staff. The average score of 5 was described as being the average performance of a first-year fellow approximately halfway through the academic year. The nurses did not know any of the fellows before their introduction to the bronchoscopy laboratory. We were unable to have blinded faculty members assess the study groups because of prior knowledge of randomization.

Statistical measuresFor comparisons between the SIM group and the control group on the solo-bronchoscopy quality outcomes, Student's two-tailed unpaired t test was used. For analysis of the SIM group's simulated bronchoscopies, one-tailed paired t tests were used to determine if there was a significant difference in performance between initial performance and subsequent performance on the simulator. A one-tailed design was used because we believed that any change in performance quality would be unidirectional (i.e., with practice on the simulator, bronchoscopists would not get worse). Linear regression was also performed on the SIM group's simulated bronchoscopies to assess the learning curve.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Observational Trial

Twenty-eight participants were enrolled in the study, with 11 novices, 8 intermediates, and 9 experts. All participants were able to complete the study. Table 1 lists the demographic data of the participants. Only the level of bronchoscopy experience was significantly related to the outcome measures. Age, computer experience, and level of video game play were not significantly related to the outcome measures. Although there were significantly fewer women in the expert category, after adjustment for bronchoscopy experience, sex was not significantly related to the outcome measures. The relatively low number of women in our study was compared with national trends in training and fellowship selection. Data from the American College of Chest Physicians regarding sex distribution among chest physicians indicated that of current members, 12,932 (85%) were men and 1,829 (12%) were women (with 543 [3%] being not reported).

The outcome measures as they relate to bronchoscopy experience level for Case 1 and Case 2 combined on the simulator are shown in Table 2. There was a significant relationship (p < 0.05) between skill level and the outcome measures procedure time, time in red-out, number of scope collisions, percentage of segments entered, and use of suction. There was not a significant relationship between experience level and amount of lidocaine used. We then sought to determine for each outcome measure which bronchoscopy experience levels resulted in significant differences between groups. For total procedure time, collisions, and percentage of segments entered, there was a significant difference between novices and intermediates/experts. For time in red-out, there was a significant difference between novices and experts.

Randomized-controlled Interventional Trial

The SIM group's experience on the bronchoscopy simulator was used to construct learning curves for bronchoscopy simulation, with bronchoscopy skill assessed by the outcomes total procedure time, number of collisions, time in red-out, and percentage of segments entered. These learning curves for bronchoscopy simulation procedure skills are shown in Figures 3 and 4. The learning curves demonstrate the rate of skill acquisition in relation to simulation experience.

View larger version (34K): [in this window] [in a new window]   Figure 3.   Learning curves for a bronchoscopy simulator as assessed by procedure time and collisions.

View larger version (30K): [in this window] [in a new window]   Figure 4.   Learning curves for a bronchoscopy simulator as assessed by percentage of segments entered and time in red-out. *p Values calculated using log10 transformation because of skewed distribution.

Total procedure time decreased from 991 s to 551 s (p = 0.006). In addition, there was a significant improvement in performance between sequential groups until the last session (attempts 16 to 20), without a clear plateau after 20 simulations. Percentage of segments entered also improved, from 90% to 97% (p = 0.002). Again, the learning curve did not demonstrate a clear plateau, although after 20 simulations greater than 95% of segments were consistently visualized. In contrast, skill acquisition in terms of learning to avoid collisions and red-out was fairly rapid. Number of collisions decreased from 110 to 70 (p = 0.04), with a plateau after approximately 10 to 15 simulations. Similarly, time in red-out decreased from 212 s to 97 s (p < 0.001), with a plateau after 15 to 20 simulations. Linear regression models for bronchoscopy score, calculated as the percentage of segments entered divided by the total procedure time, demonstrated that the bronchoscopy score increased significantly with simulator practice (p < 0.0001) with a regression coefficient of 0.00676, R² of 0.4112 (Figure 5).

View larger version (13K): [in this window] [in a new window]   Figure 5.   Linear regression analysis of bronchoscopy quality score and simulation practice.

After demonstrating learning of bronchoscopy skills for the SIM group, we went on to evaluate whether or not these learned skills would translate into actual bronchoscopy mastery. Fellows from both the SIM group and the traditionally trained group performed two actual bronchoscopies as described in the preceding protocol. No patients suffered any adverse effects, and all fellows in both groups were able to complete the protocol. The demographics and indication for bronchoscopy of the 12 patients (two per fellow) who underwent bronchoscopy are shown in Table 3. The bronchoscopy performance results are summarized in Table 4. There was a statistically significant difference between the SIM group and the control group during the actual solo-bronchoscopy as assessed by the outcomes for total procedure time, bronchoscopy quality score, qualitative assessment by a blinded bronchoscopy nurse, and amount of meperidine used.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study demonstrates the validity and potential utility of a bronchoscopy simulator. The bronchoscopy simulator, as assessed by its ability to distinguish expert from intermediate and novice bronchoscopists, was a valid and accurate reproduction of actual bronchoscopy. Performance on the bronchoscopy simulator, as measured by total procedure time, percentage of segments visualized, number of collisions, and time in red-out, correlated with bronchoscopic experience. There was a demonstrable learning curve for each of these simulated skills, with variation in the rate of skill acquisition depending upon the skill. These observations suggests that skill transfer from actual bronchoscopy to the simulation is accurate and that performance on the simulator, as measured by the foregoing criterion, may be an effective way to assess skill acquisition during bronchoscopy training.

This study also demonstrates the potential utility of simulators in teaching bronchoscopy. After 20 sessions with the simulator, despite the small sample size, there was a significant difference in terms of actual bronchoscopy performance (Table 4). The SIM group outperformed the control group with respect to bronchoscopy quality score, total time, and subjective quality as assessed by a blinded, experienced bronchoscopy nurse. We used an aggregate score, the bronchoscopy quality score, defined as the percentage of segments correctly identified divided by the amount of time taken to perform the procedure, because of the potential for bias if quality was assessed using only one measure. A quality bronchoscopy would be both fast and accurate. Thus, the ratio reflects more accurately the total quality of the bronchoscopy than either time or accuracy measures alone. A fast but inaccurate bronchoscopy is not good, whereas a tremendously slow bronchoscopy, while better than being inaccurate, is still not a quality result. The SIM group was significantly better than the control group with respect to this overall quality measure (p = 0.03).

These results should be considered in the overall context of bronchoscopy training. The amount of evidence regarding optimal bronchoscopy training is quite limited. Our current paradigm of teaching bronchoscopy has been a replication of the surgical model of teaching, with a progression from cognitive aspects of bronchoscopy to time-limited supervised bronchoscopy in low-risk patients, to more complex procedures in higher-risk patients (13). Initial patient and procedural exposure is often time-limited, recognizing the importance of patient comfort, safety, tolerance, and satisfaction with the examination. Previous methods that have been demonstrated to be of benefit in teaching bronchoscopy include instructional videos and the use of mechanical models combined with didactic teaching (14). Previous mechanical models were limited by their artificial appearance and the inability to incorporate the appearance of normal and abnormal pathology into the models (13). Animal subjects have been used to overcome this problem, but costs, ethical concerns, and anatomy differences between species limit the utility of this approach (13, 17).

Using these traditional training methods, it is not clear how best to assess competency and what constitutes an appropriate minimum level of bronchoscopy training. Even the question of how many procedures should be required to complete training is not well studied, with recommendations ranging from 25 to 100 (13, 18). Currently, the American Board of Internal Medicine does not specify a minimum number of procedures that needs to be done to learn a skill or maintain competence (3, 4). However, hospital credentials committees want standards that can be readily applied in determining privileges, and requiring a minimum number of procedures as a criterion is a frequently used method (5, 24). Evidence-based medicine has been used to establish threshold numbers of procedures needed to achieve competence in other fields, such as gastroenterology (25, 26). This has lead some investigators to recommend that threshold levels be applied for bronchoscopy training as well (24). However, there is a paucity of evidence-based medicine defining what these threshold levels should be for bronchoscopy and how competence can be directly measured.

In addition to the difficulties involved with developing evidence-based training recommendations and methods of assessing competency, there are other areas of bronchoscopy training that need to be addressed. The ongoing development of new technologies and procedures requires physicians to develop new skills. Previous studies have demonstrated that when new procedures are introduced, many physicians lack a method of receiving formal training in these new techniques, often learning them on their own (5). Developing training methods that allow physicians to acquire these new skills while minimizing patient risk during the learning process will be important. Similarly, some procedures are more complex and performed less often, making development and maintenance of expertise difficult.

The application of bronchoscopy simulation to these problems, although promising, has not been well studied. There have been no previous published reports using bronchoscopy simulators that combine virtual reality and robotic simulation, other than preliminary experiences published in abstract form (27). Indeed, a MEDLINE search of the last 15 yr using the terms "bronchoscopy," "education," "training," and "simulation" located only one citation where a new model for bronchoscopy training was developed. This involved a pig model to teach basic bronchoscopy skills, but the efficacy of the model was not compared in a randomized-controlled trial and the measurement of skill transfer from reality to the model and vice versa was not quantified (28). The use of virtual reality simulation technology has been widely accepted in other fields, but is only now gaining wider acceptance in medicine (29). Among the best studied applications in the medical field are laparoscopic simulators, cardiology patient simulators, and anesthesia simulators (10, 11, 35).

Bronchoscopy simulators may be useful for many of the problems relating to bronchoscopy training. For medical education it could be integrated into a didactic training program, providing familiarity with the instrument and allowing unlimited dedicated practice time before the first patient procedure (8). By training fellows before first patient contact, it would help to minimize the use of patients for practicing skills and ensure that trainees had some degree of skill and knowledge (9). In addition to allowing practice, the simulator could be used to measure the skill of the trainees, allowing a much more quantifiable assessment of their readiness to perform their first bronchoscopy than our current method of "see one then do one." Serial measures of simulation performance throughout the fellowship could also serve as an easy and objective measure to monitor how well fellows were progressing in terms of bronchoscopic skill.

Simulators may also play an important role in training for more complex and less common procedures. Although our study was limited in that our assessment only tested basic bronchoscopy skills, more complex skills, such as endobronchial biopsy, Wang needle biopsy, and interventional pulmonary procedures can be simulated. Future studies will need to examine whether these complex skills can also be learned more rapidly through the application of simulation technology.

Another important area that is well suited for bronchoscopy simulation is case management skills. Complications programmed into the simulation train the physician and his team to respond in a timely and appropriate manner. The most significant bronchoscopy complication, severe pulmonary hemorrhage, is rare, which limits the ability of the bronchoscopy team to practice and develop expertise in dealing with this rare life-threatening event. Simulation technology offers the opportunity to prepare for these rare but serious complications. This is analogous to pilots using flight simulators to practice their response to unexpected disasters, such as the loss of an engine or power failure in midair (10, 11, 29, 30).

Bronchoscopy simulation may also play a role in nonpulmonary physician education. For nonpulmonary physicians, generalists, and other health care providers, such as bronchoscopy nurses, respiratory therapists, and intensive care personnel, an accurate simulation may be useful in teaching airway anatomy as well as allowing a better appreciation of what is involved in a bronchoscopic examination. For medical residents, internists, and oncologists, simulation may help to clarify what bronchoscopy can and cannot do by using a "hands-on" approach. This may help improve the utilization of appropriate bronchoscopic procedures, such as TBNA for lung cancer staging (6). Thus, consumers of pulmonary services may be able to better decide when a bronchoscopic examination may be indicated. Finally, bronchoscopy simulators may aid in patient education and facilitate obtaining informed consent.

Our study provides preliminary evidence that bronchoscopy simulation will indeed prove useful for some of these applications. However, there are several possible sources of error that could have affected the results of this study, including differences between groups in terms of patient characteristics and sedation technique. Patients were randomly assigned to attempt to minimize between-group differences, and there appeared to be no significant difference in age or indications for bronchoscopy between the groups (Table 3). Although there was a significant difference in the amount of meperidine used by the SIM group, we do not believe this significantly contributed to the outcome measures. First, the study fellows controlled the amount of sedation used, not the attending physician, so any difference would still have been secondary to the fellows performing the procedure. Second, the patients in both groups demonstrated an adequate level of sedation during the procedure. The primary difference was that the control group tended to initiate sedation with 25 mg of meperidine, using additional midazolam as needed for sedation. This was the protocol that was used at the county hospital where they had initially trained. In contrast, the SIM group tended to use 50 mg of meperidine initially, with additional midazolam as needed. Thus, although the amount and type of drug used was different, the net effect in terms of sedation achieved was similar between groups and probably did not affect the outcome measures significantly.

Another limitation of this study was the small sample size necessitated by the practical limits of yearly fellowship training. This limits the conclusions that can be drawn from our investigation, and certainly larger trials will be needed to replicate these results. Other limitations include the outcome measures used. We used a 10-point scale to assess the overall quality of the bronchoscopy. A highly experienced bronchoscopy nurse had to be used, because the expert bronchoscopists had prior knowledge of randomization of the participants. Although numeric scales have been used to rate bronchoscopy training before, there are no validated or standardized instruments that have been used to measure bronchoscopy performance (4). However, we believe that the more quantitative measures of procedure time, percentage of segments correctly identified, and our bronchoscopy quality score, which was derived before execution of the study, will be more meaningful and reproducible for future studies. Other areas that need further evaluation include the long-term clinical outcome of simulation training. Although the bronchoscopy simulator appears to train fellows faster at the beginning of fellowship, it is not known how long this difference will persist, and whether or not this advantage will translate into fewer complications, more correct diagnoses, and greater patient satisfaction.

In summary, we performed a two-step validation and evaluation of a bronchoscopy simulator. Based on its ability to accurately differentiate between novice, intermediate, and expert bronchoscopists, we demonstrated that there was significant skill transfer from actual bronchoscopy to the simulator, suggesting that the simulation was accurate and that it could potentially be used for bronchoscopy competency assessment. We then evaluated the potential utility of the bronchoscopy simulator in training new pulmonary fellows and found that training with the simulator markedly improved actual bronchoscopy performance. This demonstrated that there was significant skill transfer from the simulator to actual bronchoscopy. Bronchoscopy simulation may prove to be an important tool in dealing with a variety of issues relating to initial bronchoscopy training, competency assessment, continuing physician training, and general medical education.