Providing an ideal surgical learning experience for veterinary students before graduation is a common challenge for veterinary education programs.1 The AVMA Council on Education requirements for accredited colleges of veterinary medicine (September 2016)2 specify in part that the curriculum of an accredited college shall provide instruction in the theory and practice of medicine and surgery applicable to a broad range of species. The document also states that outcomes of the DVM program must be measured, analyzed, and considered to improve the program; that schools should have a system in place to gather outcome data on recent graduates to ensure that the competencies and learning objectives in the program result in relevant entry level competencies; and that processes must be in place through which students are observed and assessed, with timely documentation to ensure accuracy of the assessment for having attained a series of competencies, including basic surgery skills and case management.2 Employers expect new graduates to be proficient in performing ovariohysterectomy and castration of dogs and cats.3–6 Ethical concerns about the use of live animals for surgical teaching have brought about the increased use of videotapes, simulation exercises, and cadavers for surgical skills training.7–10 Students acquire their surgical knowledge through didactic lectures, videotapes, or electronic (computer-based) learning and develop their technical skills through laboratories that include training models, cadavers, and live animals. Student exposure to and experience with performing elective ovariohysterectomy and castration procedures vary widely among veterinary schools,11–13 but until recently, limited information on the measurable outcomes of teaching basic surgical skills to veterinary students has been available.12–16
The relationship between practice or repetition and improvement in performance of a task is described as a learning curve.17 Learning curves for surgical procedures in human medicine are portrayed as graphical representations, with surgery time on the vertical (y) axis and experience of the surgeon on the horizontal (x) axis.18 Other outcomes used to assess surgical learning curves include complication rates, conversion from laparoscopic to open procedures, amount of intraoperative blood loss, postoperative patient outcomes, and technical skills data.17 Through the use of statistical methods and large data sets, learning or performance curves can be mapped and monitored over time.18,19 In veterinary surgery, the application of learning curve methods has recently been used to evaluate surgical times and complications. Learning curve analysis of natural orifice transluminal endoscopic ovariectomy procedures in 20 dogs revealed that approximately 10 procedures were required for an experienced team to reach an asymptotic surgery time.20 Proot and Corr21 used cumulative sum analysis to evaluate the learning curve for 1 surgeon performing tibial tuberosity advancement procedures (167 procedures in 122 dogs) and, with a 9% complication rate within 12 months after surgery, showed that the learning curve continued until the 22nd procedure. In another study22 evaluating 116 total hip replacement procedures in dogs by 1 surgeon, there were 15 failures in the first 12 weeks after surgery, and cumulative sum analysis found that approximately 44 procedures comprised the learning curve. Similar analysis methods in a retrospective study23 of 4 different surgeons performing laparoscopic ovariectomy in 618 dogs indicated a learning curve of approximately 80 procedures. Model fitting with data from 27 dogs undergoing single-port ovariectomy by an experienced laparoscopic surgeon suggested that 8 procedures would be required for a comparable surgeon to reach 90% of the optimal surgery performance (where optimal performance was represented by completion within the predicted surgery time).24
Kennedy et al14 retrospectively evaluated complications and outcomes for 513 animals (dogs and cats) that underwent ovariohysterectomy by third-year veterinary students in the teaching environment and identified 17 major complications, 15 of which involved dehiscence of the body wall incision. Median surgical times in that study14 were 105 and 180 minutes for ovariohysterectomy of cats and dogs, respectively. Burrow et al15 studied the records of 142 dogs that had ovariohysterectomy performed by fourth-year students in the veterinary hospital and found a mean surgery time of 79 minutes, with an overall complication rate of 20.6%. To the authors' knowledge, there have been no studies evaluating time for elective surgical procedures performed in sequence by veterinary medical students. The purpose of the study reported here was to determine the durations of common surgical procedures performed sequentially by fourth-year veterinary students and to use these data to statistically model an asymptotic (optimal) surgery time, estimate the number of procedures that need to be performed to reach the asymptotic time, and compare the optimal surgical times with those of an experienced veterinary practitioner.
Materials and Methods
Procedures
The protocol for the present study was classified as exempt from review by the Purdue University Institutional Review Board because individual student data would not be disclosed. Veterinary students included in the study were participants in a fourth-year elective rotation in which each student spent 2 weeks with a mobile surgery unit performing ovariohysterectomy and castration procedures and 1 week volunteering in an animal shelter.25 A maximum of 3 students performed surgery at any given time (1 student/patient) under the supervision of a shelter medicine clinician, a general practitioner with extensive clinical experience in performing and teaching spay and neuter surgery, or a board-certified veterinary surgeon. The first procedure of each type for every group of students was demonstrated by the clinician. In subsequent procedures, the students took the lead role as surgeon but were assisted by the clinician when requested by the student or when deemed necessary by the clinician for safety or teaching purposes. There was no prescribed order of procedures, except that surgeries in pediatric patients were generally performed earlier in the day to minimize the duration of food withholding and facilitate rapid return to the shelter.
Electronic medical records of elective surgical procedures (a convenience sample of ovariohysterectomies of dogs or cats and castrations of dogs) performed by fourth-year veterinary students from July 30, 2012, to May 9, 2014, were extracted from the database used for the surgery and shelter rotation program. Feline castrations were excluded because the surgical times were too short to measure accurately. Dogs and cats > 5 months and ≤ 5 months of age were classified as adult and pediatric, respectively. Medical records were included in the study when the surgery was performed by a student or an experienced practitioner (NF), the required surgical times were recorded, and the student performing the surgery completed the rotation.
All procedures were performed aseptically following the published guidelines for elective surgery on shelter animals.26 The anesthesia protocol included premedication with acepromazine maleate (0.05 mg/kg [0.02 mg/1b], SC). A combination of tiletamine hydrochloride–zolazepam hydrochloridea (100 mg/mL), butorphanol tartrateb (5 mg/mL), and dexmedetomidine hydrochloridec (250 μg/mL) was mixed in 1 vial (final concentrations for all drugs shown) and administered at a dosage of 0.03 mL/kg (0.01 mL/lb, IM) for induction of anesthesia.27 Each animal was intubated, and isofluraned (1% to 2% in oxygen) was administered as needed to maintain a surgical plane of anesthesia. Heart rate and oxygen saturation of arterial hemoglobin as measured by pulse oximetry (with the monitoring probe placed lingually) were monitored continuously; values were recorded prior to surgery and at the end of surgery. An NSAID (carprofen,e 2.2 mg/kg [1.0 mg/lb], SC; meloxicam,f 0.1 mg/kg [0.045 mg/lb], SC; or firocoxib,g 5.0 mg/kg [2.27 mg/lb], PO) was administered for postoperative analgesia. Antibiotic prophylaxis was not provided routinely; however, cefazolin (22 mg/kg [10 mg/lb], IV) was administered if there was contamination during surgery or if surgical time exceeded 90 minutes. Each surgical site was prepared aseptically and draped for surgery. The surgical techniques used varied slightly among clinicians.
Ovariohysterectomy of dogs was performed through a midline celiotomy 1 to 3 cm in length; a spay hook was used to retrieve the uterine horns, and a 2-clamp technique, with a Carmalt clamp placed on the ovarian pedicle proximal to the ovary and a Kelly forceps placed across the utero-ovarian junction, was used. Each ovarian pedicle was transected between the ovary and Carmalt clamp prior to or following ligation; ovarian pedicles in adult dogs were double-ligated, and those in pediatric dogs were single-ligated. The uterine body and its vasculature were ligated with 1 or 2 ligatures prior to transection. If excessive bleeding from the broad ligament was anticipated, it was also ligated. Size 0, 2-0, or 3-0 monofilament absorbable suture was selected according to pedicle size at the discretion of the student and clinician. At the end of the procedure, the abdominal wall incision was closed with simple interrupted absorbable monofilament sutures at each end and ≥ 1 suture in a cruciate pattern in the middle. Closure of the subcutaneous tissue was performed with a simple continuous suture pattern, and skin closure was accomplished with a simple continuous pattern in the intradermal layer, with the same suture material. Surgical glue was applied to the closed incision.
Ovariohysterectomy of cats was performed through a midline celiotomy, 1 to 2 cm in length, and was similar to the procedure in dogs except that a pedicle tie (autoligation) of each ovarian pedicle was performed in lieu of suture ligatures. A single ligature of 3-0 monofilament absorbable suture was placed around the uterine body prior to excision, and surgical wound closure was performed with 3-0 monofilament absorbable sutures in 3 layers as described for dogs. Surgical glue was applied to the closed incision.
Adult dogs were castrated through a prescrotal incision. Closed castration was performed by excision of the testis with placement of a single ligature of 2-0 to 3-0 absorbable suture on each spermatic cord, and the surgical wound was closed in 2 layers (subcutaneous and intradermal) with the same suture material in a simple continuous pattern, followed by surgical glue application. Open castration of pediatric dogs was performed through a scrotal incision near the median raphe; a figure-8 cord tie was used for spermatic cord ligation, and surgical glue was used for skin closure.
All animals received a linear tattoo with green tattoo paste (identifying them as having been neutered) immediately after closure of the incision and removal of the drapes. Surgical glue was applied to the tattoo. Patients were monitored during postoperative recovery and extubated when they were able to swallow. When they were able to lift their heads, animals were returned to the shelter for further monitoring by trained shelter personnel.
Students were required to write a surgery report for each animal, and any intraoperative complications, such as unusual anatomy, excessive bleeding, or the need to extend the incision, were noted. Follow-up by telephone, email, or text message was made to the shelter supervisors 1 to 2 days after surgery to obtain information related to the health of the animals. Animals with incisional complications that required resuturing were seen by a local veterinarian or by the supervising clinician.
Data collection
Data collected from the records of each animal included patient species, age, and sex; identity of the experienced practitioner or student who performed the procedure; time of anesthetic administration; surgery start (initial skin incision) and end (completion of tattooing, approx 60 seconds after drape removal) times; and time of extubation. Surgery time was calculated as the difference in minutes between the start and end of surgery.
Statistical analysis
Data were sorted by student, procedure, date, and time. Each surgical procedure performed by each student was numbered sequentially according to the date and time performed. Data were then grouped by procedure type (ovariohysterectomy of an adult dog, ovariohysterectomy of a pediatric dog, ovariohysterectomy of an adult cat, ovariohysterectomy of a pediatric cat, castration of an adult dog, or castration of a pediatric dog) and sequence. Surgical times for the same types of procedures performed from July 30, 2012, to October 23, 2013, by the experienced veterinarian applying the Association of Shelter Veterinarians veterinary medical care guidelines for spay-neuter programs26 were used for comparison.
For each procedure type, data for student surgeons were expressed as the least squares geometric mean surgical time and 95% confidence interval for each sequential procedure completed by ≥ 10 students. The median time for each procedure type was calculated for the experienced surgeon. The geometric mean is the arithmetic mean of the logarithmically transformed values of a variable (ie, the arithmetic mean on the log scale) followed by exponentiation of the mean to return the value to the original scale. As such, geometric mean was a more appropriate summary measure than arithmetic mean because surgical time was positively skewed (skewed to the right) in that a small number of procedures took much longer than expected. As applied in this study, the geometric mean approximates the median, but log transformation facilitates the use of parametric procedures, such as ANOVA, provided that the remaining assumptions of the statistical analysis procedure are met.
Repeated-measures ANOVA with a mixed-models procedure was applied separately for each procedure type to generate least squares geometric mean surgical time for each procedure number that accounted for the effect of student. Surgical time was log transformed for the mixed-models analysis to ensure homogeneity of variances. To describe the association between time to complete the procedure and procedure number, nonlinear and linear regression analyses were then performed by use of the least squares geometric mean values for time determined from the mixed-models procedure and the number of students contributing to each geometric mean value as a weight. Nonlinear regression followed the Levenberg-Marquardt method and used the following 3-factor exponential equation with a nonzero asymptote as described previously18:
where to is the intercept on the y-axis (initial starting point), e is a mathematical constant approximately equal to 2.71828, k is the rate of learning (unitless value), and t∞ is the surgical time at infinite procedure number under the study conditions.
This surgical time–procedure number relationship included the 3 main features of a learning curve: an initial starting point, a change in the rate of learning that is likely to be exponentially related to the number of surgical procedures performed, and an asymptotic level at which surgical performance stabilizes (optimal surgical time). Major advantages of this modeling approach were that it provided an estimate of rate of learning and the asymptotic surgical time with a 95% confidence interval for the estimate. Moreover, the 3-factor nonlinear regression model provided an estimate for the number of procedures that would have to be completed to have 95% confidence of reaching t∞. Initial estimates for to (10 to 50 minutes in increments of 10 minutes), t∞ (5 to 20 minutes in increments of 5 minutes), and k (0.1 to 0.4 in increments of 0.1) were used when applying the nonlinear regression procedure. The accuracy of the estimated values for to, t∞, and k was evaluated by determining the number of iterations required to converge to a solution, calculating the pseudo-R2 value, comparing actual versus predicted values for surgical time, and examining residual plots. The best learning curve model (nonlinear or linear) was determined on the basis of visual examination of residual plots and comparison of pseudo-R2 values (nonlinear regression) and R2 values (linear regression). The median operative times of each procedure type for the experienced practitioner were subjectively compared with 95% confidence intervals of the t∞ value for the same procedure when performed by students.
Values of P < 0.05 were considered significant for all statistical comparisons. A statistical software programh was used for all analyses.
Results
Records of 5,231 elective procedures were identified in the accessions database for the surgery and shelter rotation program during the study period. Of these, 3,056 qualified for study inclusion and were performed by student surgeons, with the total number of the specified procedures they each performed ranging from 11 to 71. A mean of 34 procedures/student was captured in the data set (Table 1). Twelve records were excluded from statistical analysis because of missing values. The median numbers of procedures performed by individual students were as follows: ovariohysterectomies of adult dogs, 6 (range, 1 to 20); ovariohysterectomies of pediatric dogs, 3 (range, 0 to 12); castrations of adult dogs, 8 (range, 1 to 19); castrations of pediatric dogs, 2 (range, 0 to 10); ovariohysterectomies of adult cats, 7 (range, 0 to 20); and ovariohysterectomies of pediatric cats, 5 (range, 0 to 19). Data for each number of procedures of a given type performed by ≥ 10 students (2,945 procedures in total) and the estimated values for variables used to define the learning curves for each procedure type were summarized (Table 2). Scatterplots of surgery time by procedure number were created to estimate the learning curves for sequentially performed surgeries completed by ≥ 10 students and to determine the number of procedures required for the modeled surgery time to be within the 95% confidence interval for the calculated optimal surgical time (Figures 1–6).
Scatterplot of surgery time by sequential procedure number for ovariohysterectomies of adult dogs performed by 88 fourth-year veterinary students during 2 weeks of a 3-week spay-neuter surgery and animal shelter rotation. Each student performed ≥ 1 of the surgeries. White circles represent individual surgery times and are slightly offset from the procedure number to improve readability. Two data points for which time was > 120 minutes are not shown. Black circles with error bars represent the geometric mean and 95% confidence interval of surgery time for each number of surgeries completed by ≥ 10 students; 568 procedures were included in these calculations. The thick linear fitted line represents the linear regression equation (surgery time = 57.5 – 2.43 × procedure number; R2 = 0.94). The mean time for the first procedure was 55 minutes; this value decreased by 2.4 minutes for each subsequent procedure up to the 11th (the highest number of these surgeries performed by ≥ 10 students).
Citation: Journal of the American Veterinary Medical Association 251, 3; 10.2460/javma.251.3.322
Scatterplot of surgery time by sequential procedure number for ovariohysterectomies of pediatric dogs performed by veterinary students. Two hundred ninety-four procedures were included in calculation of geometric mean (95% confidence interval) surgical times. The thick curvilinear fitted line represents the fitted 3-factor exponential equation (surgery time = 25 × e−0.248 × procedure number + 22), where e represents a mathematical constant approximately equal to 2.71828. The thin solid horizontal line represents the surgery time at infinite procedure number (ie, optimal surgery time). The shaded area between dashed horizontal lines indicates the 95% confidence interval for optimal surgery time. The mean time for the first procedure was 44 minutes, and the calculated asymptote was 22 minutes. Notice that it required only 2 procedures for the modeled surgery time to be within the 95% confidence interval for optimal surgery time. See Figure 1 for remainder of key.
Citation: Journal of the American Veterinary Medical Association 251, 3; 10.2460/javma.251.3.322
Scatterplot of surgery time by sequential procedure number for castrations of adult dogs performed by veterinary students. Eight data points for which time was > 60 minutes are not shown. Seven hundred twenty-one procedures were included in calculation of the geometric mean (95% confidence interval) surgery times. The thick curvilinear fitted line represents the fitted 3-factor exponential equation (surgery time = 16 × e−0.149 × procedure number + 11). Mean time for the first procedure was 25 minutes, and the calculated asymptote was 11 minutes. Notice that it required 9 procedures for the modeled surgery time to be within the 95% confidence interval for optimal surgery time. See Figures 1 and 2 for remainder of key.
Citation: Journal of the American Veterinary Medical Association 251, 3; 10.2460/javma.251.3.322
Scatterplot of surgery time by sequential procedure number for castrations of pediatric dogs performed by veterinary students. Two hundred twenty-five procedures were included in calculation of the geometric mean (95% confidence interval) surgery times. The thick curvilinear fitted line represents the fitted 3-factor exponential equation (surgery time = 41 × e−2.51 × procedure number + 8); model fit was not significant (P = 0.12), but the plot is included for comparison with other figures. Mean time for the first procedure was 12 minutes, and the calculated asymptote was 8 minutes. Notice that it required only 2 procedures for the modeled surgery time to be within the 95% confidence inter val for optimal surgery time. See Figures 1 and 2 for remainder of key.
Citation: Journal of the American Veterinary Medical Association 251, 3; 10.2460/javma.251.3.322
Scatterplot of surgery time by sequential procedure number for ovariohysterectomies of adult cats performed by veterinary students. There were 632 procedures included in calculation of the geometric mean (95% confidence interval) surgery times. The thick curvilinear fitted line represents the fitted 3-factor exponential equation (surgery time = 17 × e−0.181 × procedure number + 21). Mean time for the first procedure was 36 minutes, and the calculated asymptote was 21 minutes. Notice that it required 7 procedures for the modeled surgery time to be within the 95% confidence interval for the optimal surgery time. See Figures 1 and 2 for remainder of key.
Citation: Journal of the American Veterinary Medical Association 251, 3; 10.2460/javma.251.3.322
Scatterplot of surgery time by sequential procedure number for ovariohysterectomies of pediatric cats performed by veterinary students. There were 505 procedures included in calculation of the geometric mean (95% confidence interval) surgery times. The thick curvilinear fitted line represents the fitted 3-factor exponential equation (surgery time = 17 × e−0.259 × procedure number + 18). Mean time for the first procedure was 31 minutes, and the calculated asymptote was 18 minutes. Notice that it required 8 procedures for the modeled surgery time to be within the 95% confidence interval for the optimal surgery time. See Figures 1 and 2 for remainder of key.
Citation: Journal of the American Veterinary Medical Association 251, 3; 10.2460/javma.251.3.322
Numbers of fourth-year veterinary students completing between 1 and 20 elective surgical procedures during 2 weeks of a 3-week spay-neuter surgery and shelter rotation program from July 30, 2012, to May 9, 2014.
| No. of students completing procedures | ||||||
|---|---|---|---|---|---|---|
| Ovariohysterectomy of dogs | Castration of dogs | Ovariohysterectomy of cats | ||||
| No. of procedures | Adult | Pediatric | Adult | Pediatric | Adult | Pediatric |
| 1 | 88 | 82 | 87 | 71 | 84 | 82 |
| 2 | 87 | 64 | 87 | 58 | 83 | 76 |
| 3 | 86 | 48 | 81 | 39 | 81 | 64 |
| 4 | 78 | 40 | 76 | 26 | 76 | 55 |
| 5 | 66 | 28 | 74 | 18 | 72 | 47 |
| 6 | 52 | 21 | 67 | 13 | 60 | 43 |
| 7 | 39 | 11 | 60 | 7 | 51 | 36 |
| 8 | 27 | 6 | 49 | 4 | 40 | 25 |
| 9 | 19 | 4 | 39 | 2 | 31 | 22 |
| 10 | 14 | 2 | 32 | 1 | 25 | 18 |
| 11 | 12 | 2 | 23 | — | 16 | 14 |
| 12 | 9 | 1 | 19 | — | 13 | 13 |
| 13 | 5 | — | 15 | — | 8 | 10 |
| 14 | 2 | — | 12 | — | 4 | 5 |
| 15 | 1 | — | 8 | — | 2 | 4 |
| 16 | 1 | — | 7 | — | 2 | 3 |
| 17 | 1 | — | 4 | — | 1 | 3 |
| 18 | 1 | — | 1 | — | 1 | 2 |
| 19 | 1 | — | 1 | — | 1 | 2 |
| 20 | 1 | — | — | — | 1 | — |
Data were summarized for 590 ovariohysterectomies of adult dogs, 309 ovariohysterectomies of pediatric dogs, 742 castrations of adult dogs, 239 castrations of pediatric dogs, 652 ovariohysterectomies of adult cats, and 524 ovariohysterectomies of pediatric cats. Castrations of cats were excluded from the study. Data for another 12 surgeries were excluded because of missing information (4 ovariohysterectomies of dogs [2 adult and 2 pediatric], 4 castrations of dogs [1 adult and 3 pediatric], and 4 ovariohysterectomies of cats [all adult]).
— = Not applicable.
Estimated values for variables used to define the learning curves for a series of elective surgical procedures performed by 88 fourth-year veterinary students during 2 weeks of a 3-week spay-neuter surgery and shelter rotation program.
| Nonlinear regression | Linear regression | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Procedure type | No. of students | No. of procedures performed by ≥ 10 students | Pseudo-R2 | P value | to (min) | t∞ (95% CI; min) | k | R2 | P value | Intercept | Coefficient |
| Ovariohysterectomy of dogs | |||||||||||
| Adult | 88 | 568 | 0.940 | ND | ND | ND | ND | 0.943 | < 0.001 | 57.5 | −2.43 |
| Pediatric | 83 | 294 | 0.964 | 0.001 | 47.4 | 22.1 (4.6–39.6) | 0.248 | 0.927 | ND | ND | ND |
| Castration of dogs | |||||||||||
| Adult | 88 | 721 | 0.955 | < 0.001 | 27.0 | 10.5 (5.9–15.2) | 0.149 | 0.909 | ND | ND | ND |
| Pediatric | 73 | 225 | 0.759 | 0.12 | 49.0 | 8.3 (5.8–10.9) | 2.514 | 0.527 | ND | ND | ND |
| Ovariohysterectomy of cats | |||||||||||
| Adult | 88 | 632 | 0.957 | < 0.001 | 37.8 | 21.2 (16.4–26.0) | 0.181 | 0.901 | ND | ND | ND |
| Pediatric | 86 | 505 | 0.958 | < 0.001 | 35.4 | 18.1 (15.4–20.7) | 0.259 | 0.850 | ND | ND | ND |
Repeated-measures ANOVA with a mixed-models procedure was applied separately for each procedure type to generate least squares geometric mean values for each procedure number that accounted for the effect of student; surgery time was log-transformed for this analysis. Nonlinear and linear regression analyses were then performed by use of the least squares geometric mean values for time determined from the mixed-models procedure and the number of students contributing to each geometric mean value as a weight. Nonlinear regression used the following 3-factor exponential equation with a nonzero asymptote as described previously18: surgery time = (to–t∞) × e−(k × procedurenumber) + t∞, where to is the intercept on the y-axis (initial starting point), e is a mathematical constant approximately equal to 2.71828, k is the rate of learning, and t∞ is the surgery time at infinite procedure number under the study conditions (ie, optimal surgery time).
CI = Confidence interval. ND = Not determined.
Comparison of R2 and pseudo-R2 values for linear and nonlinear regression analysis, respectively, coupled with visual examination of residual plots indicated that the 3-factor exponential (nonlinear) equation with a nonzero asymptote provided the best description of the relationship between surgical time and procedure number for 5 of the 6 surgical procedures (Table 2). The 1 exception was the surgical time–procedure number relationship for ovariohysterectomy of adult dogs that was best described by a linear equation.
For ovariohysterectomy of adult dogs (n = 568), the (geometric) mean time required for the first procedure was 55 minutes (95% confidence interval, 40 to 78 minutes); surgery time decreased linearly with sequential performance of procedures in this category, but an asymptote was not detectable within 11 procedures (the highest number of procedures performed by ≥ 10 students; Figure 1). After 6 procedures, the mean time for this procedure was ≤ 45 minutes (95% confidence interval, 30 to 55 minutes). For 294 ovariohysterectomies of pediatric dogs, mean time for the first procedure was 44 minutes (95% confidence interval, 32 to 66 minutes) and the calculated asymptote was 22 minutes (Figure 2).
For castrations of adult dogs (n = 721), the mean surgical time was 25 minutes (95% confidence interval, 16 to 36 minutes) for the first patient, and the calculated asymptote was 11 minutes (Figure 3). Castrations of pediatric dogs (n = 225) required a mean of 12 minutes (95% confidence interval, 6 to 25 minutes) for the first patient, and the calculated asymptote was 8 minutes (Figure 4). The probability for the nonlinear regression model fit was not significant (P = 0.12) for castration of pediatric dogs; however, estimated nonlinear regression values are included in the comparison with other surgical procedures (Table 2).
Ovariohysterectomy procedures in adult cats (n = 632) required 36 minutes (95% confidence interval, 21 to 62 minutes) for the first procedure and had a calculated asymptote of 21 minutes (Figure 5). The same procedures for pediatric cats (n = 505) required a mean of 31 minutes (95% confidence interval, 20 to 45 minutes) for the first patient, and the calculated asymptote was 18 minutes (Figure 6).
For this group of 2,945 animals, all procedures were completed successfully, and there were no intraoperative deaths. There were 4 known postoperative complications; 3 animals (2 dogs and 1 cat) required additional surgery for treatment of incisional hernias. One animal died during recovery, and postmortem examination did not reveal the cause of death. Incisional complications (redness, swelling, self-trauma, or infection) were subjectively few but were not tracked specifically.
One hundred forty elective spay and neuter procedures performed by the experienced general practitioner were also studied. The median surgery times were 11 minutes for ovariohysterectomy of adult dogs (n = 48), 10 minutes for ovariohysterectomy of pediatric dogs (11), 5 minutes for castration of adult dogs (31), 4 minutes for castration of pediatric dogs (8), 6 minutes for ovariohysterectomy of adult cats (17), and 6 minutes for ovariohysterectomy of pediatric cats (25). Median surgery time for ovariohysterectomy of pediatric dogs by this practitioner was within the 95% confidence interval of the optimal surgical time (Table 2); however, the median surgery times for this individual were below the 95% confidence interval for this value (ie, times for the experienced practitioner were shorter) for ovariohysterectomy of adult and pediatric cats and for castration of adult and pediatric dogs. The median surgery time was also below the 95% confidence interval of the last (11th) ovariohysterectomy procedure performed in adult dogs by students.
Discussion
The present study was performed to attempt to map learning curves for small animal surgical procedures commonly performed by veterinary students. The initial surgery times for ovariohysterectomy and castration procedures performed by these 88 students compared favorably with previously published results.14–16,28 As expected, the learning curve analysis revealed an inverse relationship between student surgical experience and surgery time. The data were used to calculate optimal surgical times for each procedure type except for ovariohysterectomy of adult dogs: 22 minutes for ovariohysterectomy of pediatric dogs; 11 and 8 minutes for castration of adult and pediatric dogs, respectively; and 21 and 18 minutes for ovariohysterectomy of adult and pediatric cats, respectively. These optimal surgical times for students were longer than median times for the same procedure types performed by the experienced clinician (as determined by comparison with the 95% confidence interval for calculated surgical time at the infinite procedure number [ie, optimal surgical time]) except for ovariohysterectomy of pediatric dogs. Adult dogs undergoing ovariohysterectomy can vary substantially in size, conformation, stage of estrus, and body condition score. These factors, and the relatively few canine ovariohysterectomy procedures (median of 6 for adult dogs) performed by each student likely accounted for the inability to define an optimal surgical time for this patient category in the present study. The data revealed that students had to perform 2 ovariohysterectomies of pediatric dogs, 9 castrations of adult dogs, 2 castrations of pediatric dogs, 7 ovariohysterectomies of adult cats, and 8 ovariohysterectomies of pediatric cats for the modeled surgery times to be within the 95% confidence interval for the respective optimal surgical times. The reduction in time needed to complete various surgical procedures with repetition was expected, as students were expected to become familiar with the procedures, to gain more confidence, and to develop muscle memory, by which motor skills are performed with less conscious thought. In the present study, surgery time was used as 1 measure of surgical performance, considering that intraoperative complications such as inability to identify the linea alba, difficulty in retracting a uterine horn with a spay hook, slipped ligatures, dropped pedicles, or creation of an excessively long incision would increase surgery time. Other factors that can interfere with efficient performance such as inadequate depth of anesthesia or lack of instructor availability were minimized in this study by following a standard anesthesia protocol and having the supervising clinician immediately available.27,28
The present study focused on evaluating students while they performed surgical procedures in real practice as a demonstration of competence.29–31 Monitoring surgical time provides 1 quantitative measure that tracks performance.32–34 Measurable feedback is crucial to improving performance. By calculating learning curves and optimal surgical times, measurable goals for improvement can now be set for students in our program, and this provides an example for other institutions to follow in setting goals for surgical education of veterinary students.
Other means to assess student performance could include error analysis, use of global rating scales, or procedure-specific checklists.35–38 Bowlt et al16 studied competency in veterinary surgery with a combination of surgical time determination and a procedure-specific checklist. In that study,16 11 final-year veterinary students each performed 5 ovariohysterectomy procedures in dogs, and 9 of 11 students were subjectively judged as competent after performing 4 procedures.16 A surgery time of < 120 minutes was used as a criterion for success, along with successful completion of a checklist.16 Our study did not include the use of procedure-specific checklists, but the clinician monitored the successful performance of each step and gave immediate feedback to ensure that the potential for postoperative complications was minimized.
The present study has several important limitations. First, learning was characterized by reduction in surgery time alone. Prolonged surgical time is an identified risk factor for surgical site infection39,40 however, long-term information on postoperative complication rates was lacking in this study. Major complications of elective surgery are relatively uncommon,13 and postoperative complication data are not always immediately available. Each shelter involved in the student rotation program was contacted 1 to 2 days after surgery and visited every 3 to 4 weeks. Three patients were known to have postoperative complications, but additional data were not available. It is possible that some animals had dehiscence, incisional hernia, infection, or other complications that were not reported or recorded in the database.
Teaching veterinary surgery requires that the animals' needs are appropriately balanced with the students' educational needs. If an animal is severely hypothermic, unstable under anesthesia, or at risk of other major complications, assistance is provided to enable the student to complete the procedure. This so-called contingent instruction method enables faculty members to provide help when needed.34 Results of a study28 with third-year veterinary students showed that having a faculty member scrub in with students decreased overall anesthesia and surgery times for ovariohysterectomy and castration procedures in dogs. In the study reported here, faculty intervention could have impacted individual surgery times and may have reduced surgery times, especially in a student's first few cases, but this could not be investigated owing to the retrospective study design. In addition, the impact on surgery time of verbal instruction or manual assistance from the supervising clinician could not be assessed.
In the study reported here, each student performed ovariohysterectomies in dogs and cats and castrations in dogs, and we observed that the surgery times decreased with the absolute number of procedures regardless of the procedure type. This finding suggested that more experience in performing any given procedure could be required to reduce surgery times if it is the only procedure being practiced by the student. Another limitation when considering application to other academic centers is that all of the procedures in the present study were performed during 2 weeks of a 3-week rotation. If procedures are being learned or practiced over a longer time period, a greater number of repetitions might be required to improve technical skills and surgery times.35 Another consideration is that only 3 faculty members were involved in teaching these 88 students, and uniform methods and procedures were followed. The procedures were performed in our mobile unit, and close supervision was maintained.
Close examination of the scatterplots revealed a wide variation in surgical times among students for each procedure sequence. This finding was in agreement with results of another study41 that included cumulative summation methods and found a wide variation in ability among 6 students in 60 attempts at placing IV catheters in a dog mannequin thoracic limb model. The authors of that study41 identified 4 steps for catheter placement and in-control limits for demonstration of proficiency. The advantages of the method used in that study41 are that a certain degree of failed catheterization is acceptable and that individual learning curves can be generated. In our study, surgery time was used to characterize learning of common surgical procedures performed to completion because it was simple to measure. The data from a large number of students can be expressed by a mathematical function, and the resultant learning curve is a smooth line.
An extensive 2001 study18 on the statistical assessment of learning curves of health technologies resulted in a number of recommendations, including the following: the rate and length of learning and the final skill level are useful measures for describing learning, learning appears to follow a power or exponential model, the simplest models should be applied in a parsimonious manner, and the reliable assessment of learning is most likely to come from prospectively collected data for multiple operators. We therefore analyzed the surgery time–procedure number relationship with an exponential equation we developed in 201120 to model the learning curve for natural orifice transluminal endoscopic surgery in dogs. To the best of our knowledge, our 2011 study20 was the first to use this modeling approach to describe a surgical learning curve in veterinary medicine. This modeling approach was first used in human medicine in 1989 to model the cost of heart transplantation42 and in 1995 to model the learning curve for radiofrequency ablation of tachyarrhythmias.43 A minor variant of the exponential equation was used in a 2014 study evaluating the learning curve for performing laparoendoscopic single site ovariectomy in dogs.24 The exponential equation in the present study included the 3 required features of any learning curve: an initial starting point, a change in the rate of learning that is likely to be exponentially related to the number of surgical procedures performed, and an asymptotic level at which experienced surgical performance stabilizes (optimal surgical time). Major advantages of this modeling approach are that it is based on our current understanding of learning theory, is parsimonious, and directly provides an estimate for the number of procedures that must be completed to be 95% confident of reaching the calculated optimal surgical time. This information can be used to estimate the potential costs and resources required to facilitate veterinary students in becoming proficient in the procedures assessed.
An interesting finding of the present study was that median surgery times for the experienced veterinary practitioner were below the 95% confidence interval for the calculated optimal surgical time for 5 of the 6 procedures investigated. This finding provides an important methodological note for future modeling related to the basic model parameters of the learning curve, in that extensive procedural experience may result in additional gains in performance. This suggests that inclusion of an empirically observed asymptote as a prior probability distribution in a nonlinear regression model might improve model fit in data sets where the number of cumulative procedures greatly exceed the maximum modeled (14 adult canine castration procedures performed by more than 10 students) in the study reported here. The cumulative sum analysis approach modeling surgery time instead of complication rate may also provide a fruitful modeling approach for investigating the effect of extensive procedural experience on surgery time.
The nonlinear regression equation used in this study provides an estimate of the rate of change in skill development or learning (represented as k). The relative value of this variable is relevant to our understanding of skill acquisition and performance. In general terms, a higher value for k indicates a faster rate of learning. Our finding that the value for k was numerically higher for ovariohysterectomy of pediatric cats than for that of adult cats, and numerically higher for castration of pediatric dogs than for that of adult dogs, was consistent with the authors' observations that the surgical approach and the change in relative anatomic size and tissue composition (particularly the amount of fat) associated with age impacted surgery time. Future studies investigating the effect of patient obesity on the value of k would therefore be of interest in further defining the learning curve.
One question that the present study did not answer is when a veterinary student should be considered competent (or proficient) in performing elective procedures. We have only begun to understand how surgery times may indicate competent performance, and this information was gained by comparing results for 1 group of students with results of an experienced veterinary clinician who had performed > 200,000 of these procedures overall. Comparing the student data with data from 1 experienced veterinary clinician set a very high standard for optimal surgical time, and that standard should not necessarily be considered a general standard for surgical competence. With limited time and an anesthetized animal, our opinion is that in our mobile teaching laboratory, demonstrating a decrease in surgery time is 1 acceptable measure of technical skills competency when procedures are being performed under supervision and when a baseline for performance has been established. We have taken the initial steps to understand what the baseline for performance or optimal surgical times of common surgical procedures might be for our students. We strongly believe that veterinary students need ample opportunities to perform elective surgical procedures to ensure that surgical technical skills are obtained prior to graduation from veterinary school. In the present study, surgery time decreased as veterinary student experience increased for each type of elective surgery, indicating the value of repetition in improving technical skills. Further study is needed to define the optimal surgical times for ovariohysterectomy of adult dogs.
Acknowledgments
Supported in part (through the Priority 4 PAWS program) by PetSmart Charities, the Tony Stewart Foundation, the Ryan Newman Foundation, Midmark, LifeLine Mobile, Merial, Novartis, the Howard Shapiro Foundation, Fidelity Charitable Gift Fund, Dr. Robert Holmes, and Mr. Jack and Mrs. Carolyn Garrigues.
The authors declare that there were no conflicts of interest.
Footnotes
Telazol, Zoetis, Florham Park, NJ.
Torbugesic, Zoetis, Florham Park, NJ.
Dexdomitor, Zoetis, Florham Park, NJ.
IsoFlo, Abbott Animal Health, Abbott Park, Ill.
Rimadyl, Zoetis, Florham Park, NJ.
Metacam, Boehringer Ingelheim, Ridgefield, Conn.
Previcox, Merial Ltd, Duluth, Ga.
SAS, version 9.3, SAS Institute Inc, Cary, NC.
References
1. Baker HJ. The future of veterinary medical education. J Vet Med Educ 2011; 38: 319.
2. AVMA. COE accreditation policies and procedures: requirements. Available at: www.avma.org/ProfessionalDevelopment/Education/Accreditation/Colleges/Pages/coe-pp-requirements-of-accredited-college.aspx. Accessed Dec 30, 2016.
3. Greenfield CL, Johnson AL, Schaeffer DJ. Frequency of use of various procedures, skills, and areas of knowledge among veterinarians in private small animal exclusive or predominant practice and proficiency expected of new veterinary school graduates. J Am Vet Med Assoc 2004; 224: 1780–1787.
4. Greenfield CL, Johnson AL, Schaeffer DJ. Influence of demographic variables on the frequency of use of various procedures, skills, and areas of knowledge among veterinarians in private small animal exclusive or predominant practice and proficiency expected of new veterinary school graduates. J Am Vet Med Assoc 2005; 226: 38–48.
5. Clark WT, Kane L, Arnold PK, et al. Clinical skills and knowledge used by veterinary graduates during their first year in small animal practice. Aust Vet J 2002; 80: 37–40.
6. Hill LN, Smeak DD, Lord LK. Frequency of use and proficiency in performance of surgical skills expected of entry-level veterinarians by general practitioners. J Am Vet Med Assoc 2012; 240: 1345–1354.
7. Bauer MS. A survey of the use of live animals, cadavers, inanimate models, and computers in teaching veterinary surgery. J Am Vet Med Assoc 1993; 203: 1047–1051.
8. Smeak DD, Beck ML, Shaffer CA, et al. Evaluation of video tape and a simulator for instruction of basic surgical skills. Vet Surg 1991; 20: 30–36.
9. Olsen D, Bauer MS, Seim HB, et al. Evaluation of a hemostasis model for teaching basic surgical skills. Vet Surg 1996; 25: 49–58.
10. Griffon DJ, Cronin P, Kirby B, et al. Evaluation of a hemostasis model for teaching ovariohysterectomy in veterinary surgery. Vet Surg 2000; 29: 309–316.
11. Smeak DD. Teaching veterinary students using shelter animals. J Vet Med Educ 2008; 35: 26–30.
12. Gopinath D, McGreevy PD, Zuber RM, et al. Developments in undergraduate teaching of small-animal soft-tissue surgical skills at the University of Sydney. J Vet Med Educ 2012; 39: 21–29.
13. Berzon JL. Complications of elective ovariohysterectomies in the dog and cat at a teaching institution: clinical review of 853 cases. Vet Surg 1979; 8: 89–91.
14. Kennedy KC, Tamburello KR, Hardie RJ. Peri-operative morbidity associated with ovariohysterectomy performed as part of a third-year veterinary surgical-training program. J Vet Med Educ 2011; 38: 408–413.
15. Burrow R, Batchelor D, Cripps P. Complications observed during and after ovariohysterectomy of 142 bitches at a veterinary teaching hospital. Vet Rec 2005; 157: 829–833.
16. Bowlt KL, Murray JK, Herbert GL, et al. Evaluation of the expectations, learning, and competencies of surgical skills by undergraduate veterinary students performing canine ovariohysterectomies. J Small Anim Pract 2011; 52: 587–594.
17. Harrysson IJ, Cook J, Sirimanna P, et al. Systematic review of learning curves for minimally invasive abdominal surgery. A review of the methodology of data collection, depiction of outcomes, and statistical analysis. Ann Surg 2014; 260: 37–45.
18. Ramsay CR, Grant AM, Wallace SA, et al. Statistical assessment of the learning curves of health technologies. Health Technol Assess 2001; 5: 1–79.
19. Räder SB, Jørgensen E, Bech B, et al. Use of performance curves in estimating number of procedures required to achieve proficiency in coronary angiography. Catheter Cardiovasc Interv 2011; 78: 387–393.
20. Freeman L, Rahmani EY, Burgess RC, et al. Evaluation of the learning curve for natural orifice transluminal endoscopic surgery: bilateral ovariectomy in dogs. Vet Surg 2011; 40: 140–150.
21. Proot JL, Corr SA. Clinical audit for the tibial tuberosity advancement procedure: establishing the learning curve and monitoring ongoing performance for the tibial tuberosity advancement procedure using the cumulative summation technique. Vet Comp Orthop Traumatol 2013; 26: 280–284.
22. Hayes GM, Ramirez J, Langley Hobbs SJ. Use of the cumulative summation technique to quantitatively assess a surgical learning curve: canine total hip replacement. Vet Surg 2011; 40: 1–5.
23. Pope JF, Knowles TG. Retrospective analysis of the learning curve associated with laparoscopic ovariectomy in dogs and associated perioperative complication rates. Vet Surg 2014; 43: 668–677.
24. Runge JJ, Boston RC, Sharona BR, et al. Evaluation of the learning curve for a board-certified veterinary surgeon performing laparoendoscopic single-site ovariectomy in dogs. J Am Vet Med Assoc 2014; 245: 828–835.
25. Freeman LJ, Ferguson N, Litster A, et al. Service learning: Priority 4 Paws mobile surgical service for shelter animals. J Vet Med Educ 2013; 40: 389–396.
26. Looney AL, Bohling MW, Bushby PA, et al. The Association of Shelter Veterinarians veterinary medical care guidelines for spay-neuter programs. J Am Vet Med Assoc 2008; 233: 74–86.
27. Ko JC, Berman AG. Anesthesia in shelter medicine. Top Companion Anim Med 2010; 25: 92–97.
28. Freeman LJ, Huse D, Lee R, et al. Teaching veterinary anesthesia and surgery: the impact of instructor availability on anesthesia, operative, and recovery times in dogs undergoing ovariohysterectomy or castration. J Vet Med Educ 2014; 41: 390–399.
29. Miller GE. The assessment of clinical skills/competence/performance. Acad Med 1990; 65: S63–S67.
30. Rethans JJ, Norcini JJ, Baron-Maldonaldo M, et al. The relationship between competence and performance: implications for assessing practice performance. Med Educ 2002; 36: 901–909.
31. Jaffer U, Cameron AEP. Laparoscopic appendectomy: a junior trainee's learning curve. JSLS 2008; 12: 288–291.
32. Duclos A, Carty MJ, Peix JL, et al; CATHY Study Group. Development of a charting method to monitor the individual performance of surgeons at the beginning of their career. PLoS One 2012; 7: e41944.
33. Wood D, Wood H, Ainsworth S, et al. On becoming a tutor: toward an ontogenetic model. Cognit Instruct 1995; 13: 565–581.
34. Sadideen H, Kneebone R. Practical skills teaching in contemporary surgical education: how can educational theory be applied to promote effective learning? Am J Surg 2012; 204: 396–401.
35. Ericsson KA. Deliberate practice and the acquisition and maintenance of expert performance in medicine and related domains. Acad Med 2004; 79: S70–S81.
36. Hardie EM. Current methods in use for assessing clinical competencies: what works? J Vet Med Educ 2008; 35: 359–368.
37. Maan ZN, Maan IN, Darzi AW, et al. Systematic review of predictors of surgical performance. Br J Surg 2012; 99: 1610–1621.
38. Reznick RK, MacRae H. Teaching surgical skills—changes in the wind. N Engl J Med 2008; 256: 2664–2669.
39. Beal MW, Brown DC, Shofer FS. The effects of perioperative hypothermia and the duration of anesthesia on postoperative wound infection rate in clean wounds: a retrospective study. Vet Surg 2000; 29: 123–127.
40. Eugster S, Schawalder P, Gaschen F, et al. A prospective study of postoperative surgical site infections in dogs and cats. Vet Surg 2004; 33: 542–550.
41. Campbell RD, Hecker KG, Biau DJ, et al. Student attainment of proficiency in a clinical skill: the assessment of individual learning curves. PLoS One 2014; 9: e88526.
42. Smith DB, Larsson JL. The impact of learning on cost: the case of heart transplantation. Hosp Health Serv Adm 1989; 34: 85–97.
43. Danford DA, Kugler JD, Deal B, et al. The learning curve for radiofrequency ablation of tachyarrhythmias in pediatric patients. Am J Cardiol 1995; 75: 587–590.
