Abstract
Objectives
The purpose of this study was to evaluate the accuracy of visual weight estimation by veterinary students for free-roaming cats presented in traps during trap–neuter–return (TNR) events and identify variables that predict the percentage difference between estimated and actual weights.
Methods
During five TNR events, veterinary students, veterinarians and technicians visually estimated weights for 308 cats. Actual cat weights were measured, and the accuracy of visual estimates was evaluated using the percentage of estimates within 10% (PW10) and 20% (PW20) of actual weight. Predictors of percentage difference were analyzed using mixed-effects linear regression. Dosing accuracy of the induction cocktail was assessed by comparing estimate-derived drug doses to the target range.
Results
Student estimates achieved a PW20 accuracy of 44%, lower than veterinarians (56%) but higher than technicians (35%). Accuracy within 10% (PW10) was limited across all groups, with students achieving 14% compared with veterinarians at 33%. Dosing based on student estimates fell within the target range for 85% of cases, compared with 95–96% for veterinarians, technicians and consensus estimates. Systematic errors included overestimating lighter cats and underestimating heavier cats, with posture and using kilograms vs pounds also affecting accuracy.
Conclusions and relevance
Although student estimates were less accurate than those by veterinarians, dosing derived from these estimates was clinically acceptable in most cases. Training on systematic biases and leveraging consensus estimates may improve accuracy. Integrating veterinary student weight estimates into TNR programs is feasible with appropriate safeguards, including training on systematic estimate biases, estimate consensus and post-induction monitoring.
Plain language summary
When free-roaming cats are brought to trap–neuter–return (TNR) clinics for spay/neuter surgery, they stay in covered traps for safety. Because they cannot be weighed before anesthesia, veterinarians and technicians usually guess their weight to decide how much medicine to give. This study looked at how accurately veterinary students could make these weight guesses and whether their estimates were safe to use. During five TNR events, veterinary students, veterinarians and technicians estimated the weights of 308 cats. The cats’ true weights were recorded after anesthesia. Researchers compared the guesses with the actual weights and checked whether the guessed weights would have resulted in safe anesthesia doses. Students’ weight guesses were less accurate than veterinarians’ but better than those of technicians. Only 14% of student guesses were within 10% of the real weight, and 44% were within 20%. Students tended to overestimate the weight of small cats and underestimate larger ones. Despite this, 85% of doses calculated from student guesses still fell into the target dosing range, compared with 95–96% for veterinarians, technicians and consensus estimates. Factors such as whether weight was given in pounds or kilograms and the cat’s posture influenced accuracy. The study shows that although veterinary students’ estimates are not as precise as those of professionals, they usually provide safe doses. With extra training about common errors, using pounds instead of kilograms and cross-checking estimates with others, students’ weight guesses could be safely integrated into TNR programs. This approach may reduce cat stress by avoiding repeated trap cover lifting and help students gain valuable clinical experience, as long as safeguards like careful monitoring and use of reversible anesthesia are in place.
Keywords
Introduction
Free-roaming cats are presented to trap–neuter–return (TNR) programs for sterilization in covered humane traps of various sizes and models. These cats are typically unsocialized to humans 1 and cannot safely be removed from the trap and weighed before the induction of anesthesia. Visual estimation of the cat’s weight is one way of generating the dose of anesthetic induction agents. Accurate weight estimation is critical for ensuring the safety of the cats and veterinary personnel. Inaccurate dosing can lead to underdosing, resulting in the cat regaining consciousness prematurely, or overdosing, which may cause adverse effects such as bradycardia, bradypnea and even death. 2 Cats appear to be at greater risk from error compared with dogs, perhaps due to lower body weight or physiologic differences. 3 Medical errors that compromise the health of the cats can negatively impact the morale of veterinarians and students. 4
Visual estimates have been evaluated for different species, including cats,5,6 dairy cattle, 7 dogs,6,8 horses 9 and humans. 10 Although there is no gold-standard goal of accuracy, a systematic review of body weight estimation from measurements suggested a minimum of 70% of estimations within 10% of actual weight (PW10) and 95% within 20% (PW20) for an estimation system to be considered accurate. 11 A study of visual weight estimation in cats presented to an emergency department (ED) showed that PW20 was in the range of 63–66%, while PW10 was in the range of 33–50% for veterinary professionals. 5
For TNR events hosted by Midwestern University’s College of Veterinary Medicine (MWU CVM), the initial visual examination to grossly assess a cat’s health status and check for an ear tip is typically performed by veterinary students. Separately, a veterinarian or technician estimates the weight to determine the dose of anesthetic induction cocktail. The visual examination is stressful for cats as it requires the trap cover that is providing a hiding space to be lifted. 12 Stress to the cats may be reduced by having veterinary students generate the weight estimate during the initial visual examination; however, the accuracy of visual weight estimation by veterinary students for cats in traps is unknown. This study aimed to evaluate the accuracy of visual weight estimates by veterinary students during TNR events and, secondarily, to identify variables that systematically predict deviations between estimated and actual weights.
Materials and methods
This study was conducted during five TNR events at MWU CVM between October 2021 and May 2022, with 77–101 cats per clinic day. Any cat presented for TNR services was eligible for inclusion in the study. Cats for which a student estimate could not be obtained, or which did not have an actual weight recorded, were excluded. Volunteer veterinary students of all years were involved with the clinic. A pre-event online training module and orientation that stressed patient safety each morning were mandatory. The students were overseen by veterinarians and performed nearly all clinical duties, including pre-anesthetic visual examination, induction of anesthesia, post-anesthetic examination and preparation of the patient, aseptic preparation, sterilization surgery and recovery. However, the induction dose of anesthetic was determined based on a visual weight estimation (pounds) provided by a veterinarian or technician and recorded in the medical record (official weight estimate). The actual dose of anesthetic induction cocktail was the weight estimate multiplied by 0.01 ml, which was rounded upwards if it was fractional. The official weight estimate was also used to determine whether cats appeared to fall below the weight limit of 1.5 pounds required to undergo surgery. The actual weight of the cat was measured in kilograms through use of a baby scale (Brecknell MS-15) after induction of anesthesia.
Permission for human subject research was granted from the Midwestern University Institutional Review Board (IRBAZ-5113). After obtaining informed consent, researchers asked the veterinary students, veterinarians and technicians to provide a visual weight estimate. Estimates were made anonymously in writing and could be given in pounds or kilograms. Because of the ratio of cats to volunteers, student participants were assigned a temporary code for the day as they could provide a weight estimate for more than one cat. Veterinarians and technicians were provided with a code for the duration of the study and submitted weight estimates at their convenience. Although the same group of veterinarians and technicians provided both the official weight estimate used to calculate the anesthetic dose and the study estimate, the official weight estimate in the medical record was determined separately from the study estimate. Cat posture (standing, sitting, crouching, lying) at the time of each estimate was recorded alongside coat color, coat length (short, medium, long), coat pattern (patterned or solid) and actual weight. The induction dose, number of doses administered, sex, estimated age, body condition score (BCS), pregnancy status, use of pounds vs kilograms, use of whole numbers and the official weight estimate were also noted.
The anesthetic protocol was a pre-mixed cocktail created by reconstituting tiletamine/zolazepam to a concentration of 100 mg/ml using 2.5 ml dexmedetomidine (0.5 mg/ml) and 2.5 ml of butorphanol (10 mg/ml). 13 This protocol is dosed at 0.02–0.04 ml/kg and is considered to have a wide range of safety and efficacy. 13 A dose of 0.022 ml/kg (0.01 ml/lb) was used for induction of anesthesia, and additional doses were administered as needed to ensure that cats achieved a suitable stage of anesthesia (completion of stage II) 14 for safe removal from the trap and completion of surgical preparation. Technicians and a small group of student clinic coordinators, who had received additional training, induced anesthesia through the trap using a trap separator to briefly immobilize the cat for an intramuscular injection delivered via an insulin syringe. The identity of the individual administering the injection was not recorded. Once the cats were injected, they were continuously monitored by veterinary students to ensure they were breathing and to determine when they had completed stage II of anesthesia. 14 This stage was initially determined by a lack of spontaneous musculoskeletal movement and response to external stimulation, such as opening the trap door. Depth of anesthesia after removal from the trap was monitored via palpebral reflex, jaw tone, limb muscle tone and respiratory rate. 14 A supraglottic device (V-Gel; Docsinnovent) was placed once cats were removed from the trap. Anesthesia was maintained via isoflurane during surgery. The analgesic protocol consisted of robenacoxib (2 mg/kg SC) for cats estimated to be aged 4 months or older, and buprenorphine (0.02 mg/kg SC) for kittens estimated to be aged less than 4 months.
Statistical methods
Commercial statistical software (Stata 18; StataCorp) was used for analysis. Data normality was assessed by tests of skewness and kurtosis. Normal data were presented as mean and SD, and non-normally distributed data were presented as median and interquartile range (IQR), expressed as quartiles 1–3. Estimates in pounds were converted to kilograms by multiplying by 0.453592. Percentage difference was calculated as ([estimated weight in kg – actual weight] / actual weight in kg × 100). A consensus estimate was determined from the mean of student and technician estimates. Veterinary estimates were not included in the consensus estimate since veterinarians are not routinely directly involved in the induction process. The accuracy of visual weight estimates was determined by PW20 and PW10, as well as the percentage within the dosing range. Proportions were compared using pairwise two-sided tests of proportion, while variance was compared using pairwise two-sample variance-comparison tests. Mixed-effects linear regression (MELR) with date and student code as random effects was used to assess the impact of the captured variables on the percentage of difference. Models were created using a combination of backwards stepwise design and plausibility with competing models evaluated based on Akaike Information Criterion and Bayesian Information Criterion. Models were validated by visual inspection of the residuals. P <0.05 was considered statistically significant.
Results
Cats
A total of 315 cats met the inclusion criteria for the study. Of them, five were excluded because of a lack of a student weight estimate and two were excluded for not having an actual weight, leaving a total of 308 cats enrolled in the study (Table 1). No cats died during the duration of the study.
Demographic variables for the 308 cats enrolled in the study
Data are n (%), mean ± SD or median (interquartile range [expressed as quartiles 1–3)]
Estimators
There were 108 unique students as defined by the date and student code. Of the 308 student estimates from three different clinic dates, seven were missing a code. Each student provided a median of two estimates (IQR 2–3, range 1–10). Students preferred to guess in pounds, with 87/108 (81%) unique students providing an estimate in pounds instead of kilograms. Of these 108 students, 51 (47%) were first-years, 33 (31%) were second-years, 12 (16%) were third-years, 11 (10%) were fourth-years and one was missing a year specification. For future career track, 80 (74%) were small animal, one (1%) was large animal, 21 (19%) were mixed and six (6%) were undecided. Students reported a median of 2 years (IQR 1–4) of experience working in a veterinary practice and had previously volunteered for a median of one TNR (IQR 0–5). In addition to the students, two technicians and three veterinarians provided estimates. Of the 308 cats with a student estimator, 141 had an estimate from one of the three veterinarians, and 194 had an estimate from one of the two veterinary technicians. Six cats had only a student estimate. Most estimates were reported as whole numbers, with 71% (219/308) of students, 82% (116/141) of veterinarians and 98% (191/194) of technician estimates falling into this category.
Accuracy
Students had a mean percentage difference from the actual weight of 17% (± 44), with a PW20 of 44% and a PW10 of 14%. The mean percentage difference compared favorably to that of other estimators, but students exhibited greater variance and the lowest PW10 (Table 2). All estimators tended to overestimate lighter cats and underestimate heavier cats (Figure 1), with an inflection point at approximately 4 kg.
Percentage difference with SD, percentage of estimates within 20% of actual weight (PW20) and PW10 (percentage of estimates within 10% of actual weight) for students, veterinarians, technicians, consensus (average of students and technicians) and official estimate from the medical record (made by veterinarian or technician)
P value is as compared with student estimates. Significant differences (P < 0.05) appear in bold
CI = confidence interval
Scatterplots of the percentage difference ([estimated weight – actual weight] / actual weight × 100) and actual weight for estimates made by students, veterinarians, technicians, consensus (average of students and technicians) and official overlaid by Lowess line. Green squares = estimates within 10% of actual, blue diamonds = estimates within 20% of actual and red circles = estimates more than 10% from actual. Horizontal line at 0 and vertical line at point at which the Lowess line crosses 0 where estimates tend to shift from overestimating to underestimating
Accuracy within dosing range
From the medical record, the median actual first dose administered was 0.03 ml/kg (IQR 0.02–0.03, range 0.01–0.05), which was within the target dose range of 0.02–0.04 ml/kg for 302/307 (98%) cats that were induced. For the five cats with an actual dose outside the dosing range, three were underdosed at 0.01 ml/kg and two overdosed at 0.05 ml/kg. The student weight estimate would have generated a dose within the dosing range (Figure 2) for 85% of the estimates (261/308), with 29 outside the range being underdosed at 0.01 ml/kg and 18 overdosed at 0.05–0.07 ml/kg. The percentage of doses calculated to be within dosing range for students was lower compared with the percentage for veterinarians, technicians, consensus and the official estimate, although the median dose was similar and the IQR was identical (Table 3).
Histograms of the dose (in ml/kg) that were (actual) or would have been administered to cats based on the respective weight estimate (student, veterinarian, technician, consensus between student and technician, and official). Vertical lines at 0.015 and 0.044 indicating the target dose range after rounding to the nearest 0.01
Median, interquartile range (IQR), range and percentage within target dose range (0.02–0.04 ml/kg) for doses calculated based on the weight estimates of students and veterinary professionals
Predictors of percentage difference
In multivariable MELR, increasing cat weight was associated with a percentage difference of −13% (95% confidence interval [CI] −17 to −10; P <0.0001), while estimating in kilograms rather than pounds was associated with a percentage difference of 19% (95% CI 3–35; P <0.019). Compared with cats that were standing, those that were sitting had a percentage difference of −32% (95% CI −55 to −9; P = 0.007), and those that were lying or crouching had a percentage difference of −23% (95% CI −44 to −0.7; P = 0.043) (Figure 3). No student characteristic – including duration of experience as a technician before veterinary school (P = 0.566), year in school (P ⩾0.423), career track (P ⩾0.218) or number of previous TNRs (P = 0.805) – was significant. The intercept represented a percentage difference (overestimate) of 75%, indicating that variables with negative values improved the accuracy of estimates, while those with positive values decreased accuracy. Cats under 4 kg tended to be underestimated by an average of 24%, whereas cats over 4 kg tended to be overestimated. The intraclass correlation coefficient (ICC) was 0.5, suggesting that approximately 50% of the variance in the outcome was explained by differences among student estimators.
Box plot of percentage difference between actual and students’ estimated weight by the cats’ posture at the time of the estimate
Number of induction doses
Of the 307 cats that were induced with the anesthetic premix, 243 (79%) received one dose, 53 (17%) two doses and 11 (4%) three doses. Compared with cats receiving one dose (Figure 4), there was no difference in weight for cats receiving two doses (P = 0.857); however, cats with three doses tended to be a mean of 0.8 kg heavier (P = 0.014) compared with cats that received one or two doses. The initial dose was a median of 0.03 ml/kg (IQR 0.02–0.03. range 0.01–0.04) for cats requiring more than one dose. All cats requiring three doses had an initial dose in the range of 0.02–0.03 ml/kg.
Box plot of the number of injections required to induce anesthesia by actual weight
Discussion
This study evaluated the accuracy of visual weight estimation by veterinary students for free-roaming cats presented in traps during TNR events. Student estimates achieved a PW20 accuracy of 44%, which was lower than veterinarians (56%) but comparable to the consensus estimate (44%) and official estimates (46%). Although accuracy within 10% of actual weight was limited across all groups, the majority (85%) of weight-based dosing calculations derived from student estimates fell within the clinically acceptable dosing range of 0.02–0.04 ml/kg. Predictors of estimation percentage difference included cat weight, posture and whether the weight was estimated in kilograms or pounds, with systematic tendencies to overestimate lighter cats and underestimate heavier cats.
In a study of cat weights visually estimated in an ED, 8 visual weight estimates generated by a mix of veterinarians, veterinary technicians and veterinary students had a PW10 of 20%, which was similar to the PW10 range observed in the present study (14–33%), suggesting that weight estimation may not be impacted by cats being in traps. Student performance in that study could not be directly compared with this one, as only 4/28 cat weight estimates in the ED were from students. As in this study, overestimation of cat weight was more common than underestimation; however, they found a non-significant trend for overestimation rather than underestimation as weight increased, and that weight was overestimated as BCS increased. Dogs in the study with longer hair had overestimated weights when they were a lower weight and underestimated weights when they were a higher weight. Unlike cats, weight estimates tended to underestimate the weight of dogs with higher BCS. Neither BCS nor hair length was significant for the cats in this study.
In another study of cat weight visually estimated by owners, veterinary technicians, house officers and attending clinicians in an ED, 5 the PW20 was in the range of 63–73% while the PW10 was in the range of 33–59%. In comparison, we found a range of 35–56% for PW20 and 14–33% for PW10. Although all estimators in the ED study were more likely to overestimate the cat’s body weight, it is unclear whether this was because of the pattern we observed in the present study, with overestimates of lighter cats and underestimates of heavier cats, as only the absolute percentage error was analyzed. On average, their cats were heavier, at a mean of 4.9 kg compared with 2.8 kg in the present study, which would put them closer to the inflection point that we observed at approximately 4 kg. Unlike this study, which did not see an improvement in student accuracy with duration of experience in the veterinary field or year in school, there was a weak negative correlation between the length of experience and absolute percentage error for the veterinary professionals.
For a study limited to dogs, 6 PW10 for veterinary technicians, house officers and attending clinicians was in the range of 35–41%, while the PW20 was in the range of 60–70%. This is higher than that observed in this study (35–56%), but this may be due to the constrained denominator for the weight of cats. The maximum weight observed in this study was 5.5 kg (12 lbs); over- or underestimating by one unit would create a percentage difference of 18–100% for kilograms or 8–50% for pounds. For a 25 kg (55 pound) dog, a one unit difference would generate a percentage difference of 4% (2% for pounds). In addition, many of the dogs in the study were carried by the people estimating their weight, although this did not make a difference in the study involving only cats. 5 Similar to the present study, veterinarians demonstrated the greatest accuracy as measured by PW10 and PW20, and the length of veterinary experience did not affect the accuracy of the estimates.
Increasing accuracy
Student estimates tended to be less accurate, as measured by PW10 and PW20, than other estimators, although their mean percentage difference was lower than all others except veterinarians. However, students had greater variance, generating more extremely wrong estimates than other estimators, which was reflected in the lower percentage within the dosing range. The relatively high ICC for student estimates (0.5) indicates that some students are better at estimating than others, suggesting that there may be an opportunity to increase the accuracy of less-accurate students using the systematic predictors noted in the present study. Using pounds instead of kilograms for students comfortable with both units would increase the accuracy, potentially because pounds are more granular, which is relevant in a relatively constrained weight range of 1.1–5.5 kg. The use of fractional numbers, particularly for estimates in kilograms, would theoretically increase accuracy, but this was not observed.
Use of student estimates clinically
Although student-derived estimates had lower accuracy than those of veterinarians, it may be possible to implement student estimates responsibly. In addition to increasing the accuracy of estimates through education on systematic bias, student estimates can be ‘double-checked’ by the technicians who induce anesthesia just before induction, which would not require an additional removal of the trap cover (the consensus estimate). Although a greater percentage of difference was observed for consensus (25%) than student (17%) estimates, the consensus estimate had lower variation, with an SD of 32 vs 44. Dosing at the lower end of the drug protocol resulted in more doses outside the target range being underdoses than overdoses, despite students tending to overestimate cat weights. Vigilant monitoring of cats after induction, combined with the use of a reversible drug protocol, helps protect against adverse effects associated with anesthetic overdose.
Limitations
To ensure anonymity, student participants were assigned a temporary code that was tracked only for the duration of the clinic. Some students may have volunteered for multiple clinics during the study period, which was not accounted for statistically. The limited number of veterinarians and technicians who participated in weight estimation limits the generalizability of the findings for those professional groups. All students were from the same institution, and their prior clinical experience may not reflect that of veterinary students elsewhere. However, neither years of technician experience before veterinary school nor year of study were significant predictors of percentage difference. The individual who induced anesthesia in each cat was not recorded, so variation between inducers could not be considered when evaluating the number of injections. Adverse events other than mortality were not tracked.
Conclusions
This study highlights both the challenges and potential of visual weight estimation by veterinary students for cats presented in traps. Although students demonstrated lower accuracy – as measured by the proportion of estimates within 10% and 20% of actual weight – compared with veterinarians, their estimates resulted in dosing calculations within clinically acceptable ranges for most cases. Identified predictors, including cat weight, posture and the use of pounds vs kilograms, suggest that estimation accuracy could be improved through targeted education. Student weight estimates could likely be used safely within a structured TNR program, provided appropriate safeguards are in place, such as training in weight estimation, consensus-based estimation, use of a reversible drug protocol with a wide safety margin and vigilant post-induction monitoring.
Footnotes
Conflict of interest
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Ethical approval
The work described in this manuscript involved the use of non-experimental (owned or unowned) animals. Established internationally recognized high standards (‘best practice’) of veterinary clinical care for the individual patient were always followed and/or this work involved the use of cadavers. Ethical approval from a committee was therefore not specifically required for publication in JFMS. Although not required, where ethical approval was still obtained, it is stated in the manuscript.
Informed consent
Informed consent (verbal or written) was obtained from the owner or legal custodian of all animal(s) described in this work (experimental or non-experimental animals, including cadavers, tissues and samples) for all procedure(s) undertaken (prospective or retrospective studies). For any animals or people individually identifiable within this publication, informed consent (verbal or written) for their use in the publication was obtained from the people involved.
