Skip to main content
NCBI home page
Veterinary Medicine and Science logoLink to Veterinary Medicine and Science
. 2025 Jul 18;11(4):e70496. doi: 10.1002/vms3.70496

Evaluating the Reliability and Accuracy of Senior Veterinary Students in Detecting and Scoring Lameness in Dairy Cows

Yalcin Alper Ozturan 1,✉, Ibrahim Akin 1
PMCID: PMC12271826  PMID: 40678867

ABSTRACT

Background

Lameness detection is essential for effective dairy cattle management, with accurate diagnosis improving animal welfare and reducing economic losses. Senior veterinary students must acquire these skills before graduation to ensure competent diagnosis in the field.

Objectives

This study aimed to evaluate the reliability and accuracy of senior veterinary students in detecting and scoring lameness in dairy cows.

Methods

The study included 201 senior veterinary students who scored lameness in cows using video recordings and a 5‐point scoring system. Students’ lameness scores were compared to those assigned by an experienced observer using a confusion matrix, with sensitivity, specificity, and accuracy calculated. Intra‐rater reliability was assessed using intraclass correlation coefficients, while inter‐rater reliability was evaluated using Krippendorff's alpha. Binary logistic regression was performed to assess the impact of lameness severity on detection accuracy.

Results

Students demonstrated high accuracy for severe lameness (93.67%) and healthy cases (85.93%), with sensitivities of 75.84% and 74.46%, respectively. However, sensitivity for mild to moderate lameness was lowest. Specificity ranged from 81.87% for mild cases to 98.12% for severe cases. Inter‐ and intra‐rater reliability showed various agreement coefficients across lameness categories. Logistic regression indicated decreased accuracy with increasing lameness severity.

Conclusion

Gaps in detecting intermediate lameness highlight the need for enhanced training methods in veterinary education. Integrating advanced tools can improve diagnostic accuracy and support better lameness detection in practice.

Keywords: dairy cattle, lameness, learning, observer reliability, veterinary education

Senior veterinary students assessed lameness in dairy cows using video footage and a 5‐point scoring system. The study evaluated their diagnostic accuracy and reliability, highlighting low to moderate agreement and the influence of lameness severity on detection accuracy, with implications for veterinary education and clinical training.

graphic file with name VMS3-11-e70496-g001.jpg

1. Introduction

Lameness in dairy cows significantly affects both animal welfare and production economics, associated with reduced milk yield, treatment expenses, impaired reproductive performance, and increased risks of culling or premature herd removal. (Huxley 2013; Dolecheck and Bewley 2018; Akin and Akin 2018) Lameness detection primarily relies on manual locomotion scoring through observational assessments. Although automated lameness detection methods have been introduced, their widespread adoption remains limited due to high installation costs and system complexity. (Flower and Weary 2009; Alsaaod et al. 2019; Kang et al. 2021; Akin et al. 2024) As a result, manual scoring remains the gold standard. (Sprecher et al. 1997; O'Callaghan et al. 2003; Espejo et al. 2006; Flower and Weary 2006; Lambertz 2019) However, this approach is heavily dependent on the observer's expertise, making it subjective and prone to inaccuracies, particularly in the early detection of lameness, which may lead to delayed treatment and economic losses. (Leach et al. 2013; Renn et al. 2014; Thomas et al. 2016; Kang et al. 2021) The ability to accurately distinguish between lame and non‐lame cows often relies on the observer's experience and perceptual skills. This underscores the critical role of practice and veterinary education in addressing these challenges effectively.

Veterinary students develop lameness assessment skills through a combination of lectures, textbooks, and hands‐on clinical exposure during rotations and farm visits. (Bell 2015; Weaver et al. 2018) At Aydin Adnan Menderes University, Faculty of Veterinary Medicine, decision‐making skills such as clinical rotations, farm visits and hands‐on training are introduced in the third year of the veterinary curriculum and last till the last year of the degree. Furthermore, the faculty holds international and national accreditation standards of the European Association of Establishments for Veterinary Education (EAEVE) and the National Association for Evaluation and Accreditation of Veterinary Medicine Educational Institutions and Programs (VEDEK). Despite this structured training, even veterinarians often face difficulties in detecting mild lameness cases. (Brenninkmeyer et al. 2007; Thomsen et al. 2008; Tunstall et al. 2021) Gait assessment skills have been shown to improve with experience, as both intraobserver (the same individual assessing the same scenario multiple times) and interobserver (different individuals assessing the same scenario) agreements increase over time. (Main et al. 2000; Brunnekreef et al. 2005; Schieder et al. 2020) However, assessments may still result in consistent errors, even as individual judgements stabilise. (Kawamura et al. 2007; Waxman et al. 2008) Studies have explored the experience‐dependent development of gait assessment abilities across various species, including cows along with experienced veterinarians (Brenninkmeyer et al. 2007; Thomsen et al. 2008; Tunstall et al. 2021), humans (Brunnekreef et al. 2005; Read et al. 2003; Williams et al. 2009), dogs (Waxman et al. 2008), horses (Keegan et al. 1998) and pigs (Main et al. 2000). Practical training sessions, guided by institutional experts, remain a cornerstone in developing these critical skills in veterinary education.

Recent advancements in teaching methods have integrated video‐assisted tools and computer simulations as supplementary resources for lameness evaluation training. (Barstow et al. 2014; Starke and May 2017; Schieder et al. 2020) These tools offer standardised training opportunities, independent of clinical caseloads, enabling self‐directed learning and enhancing perceptual accuracy. Active learning approaches, which combine cognitive engagement with physical interaction, have been shown to improve knowledge retention and skill acquisition. For example, combining lectures with visual and physical demonstrations significantly increases retention compared to lectures alone. (Dahmer 2007) Moreover, physical self‐experience, such as mimicking movement patterns associated with pain or dysfunction, fosters a deeper understanding of locomotor changes resulting from lameness. (Geber and Wilson 2010) While video‐based training tools, hands‐on training, and clinical rotations have been incorporated into curricula in faculties endorsed by national and international organisations, their outcomes and impacts, particularly for senior students about to start their careers, have not been assessed. For example, video‐based training tools have been effectively applied in equine medicine (Barstow et al. 2014), yet their impact in bovine lameness evaluation remains unexplored. This gap highlights the need for studies focused on how veterinary students develop lameness detection skills in dairy cows. Such research could inform the development of systematic teaching methods aimed at improving diagnostic accuracy and reliability in practice.

The aim of this study was to evaluate the accuracy and reliability, including both intra‐rater and inter‐rater agreement, of senior veterinary students in detecting and scoring lameness in dairy cows following theoretical education, video‐assisted training, hands‐on practice, clinical rotations, and farm visits. We hypothesized that students would be more accurate in identifying cows with either no lameness or severe lameness compared to intermediate levels, and that detection accuracy would decline non‐linearly as lameness severity shifted away from these extremes. To test these hypotheses, binary logistic regression was used to examine the effect of lameness severity on detection accuracy, alongside confusion matrix analysis. The study also aimed to identify challenges linked to different lameness severities and to explore strategies for enhancing veterinary educational methods.

2. Materials and Methods

The authors confirm that the ethical policies of the journal, as noted on the journal's author guidelines page, have been adhered to and the appropriate ethical review committee approval has been received.

2.1. Animals and Video Acquisition

Video recordings used in this study were obtained from a commercial dairy farm with a herd size of 250 cows located in Aydin province of Turkiye. The farm owner provided informed consent, and all recordings were taken from the milking alley. The study included lactating cows of varying lactation numbers (average lactation: 3) and lactation stages (average: mid‐lactation, 120–240 days in milk). Observations were made as cows exited the milking parlour through a narrow corridor permitting only one cow to pass at a time. To ensure anonymity, farm personnel were instructed to position cows' collar tags on the side opposite the camera before video recording. A digital video camera (Nikon D3200) was mounted on a tripod placed 5 metres from the corridor and positioned 1.35 metres above the floor. The gait of each cow was recorded while walking in a straight line on a flat concrete surface. Videos were taken from a side view using a tripod‐mounted digital camera, positioned to capture the full body of the animal. Recordings were captured at a resolution of 1080 × 1920 with a frame rate of 30 frames per second.

2.2. Video Processing

All collected videos were systematically reviewed by the researchers to verify consistency between live observations and video recordings. All cows in the videos were assessed using the Sprecher's 5‐point lameness scoring (LS) system, where LS1 indicated healthy cows, and LS2 to LS5 represented varying degrees of lameness: mild lame, moderately lame, and severe lame, respectively. (Sprecher et al. 1997) The assessments were conducted by a researcher with a Ph.D. in bovine lameness and 20 years of experience in the field. Based on these reviews, 15 cows (LS1, LS2, LS3, LS4, and LS5; n = 3 per score) were systematically selected to meet three key methodological requirements: (1) statistical power for reliability analysis through triplicate evaluations of each severity level, (2) comprehensive coverage of the full lameness spectrum without overwhelming participants during scoring sessions, and (3) practical feasibility of video processing and scoring within the study timeframe. Corresponding video segments, each 20 s long and featuring one cow per video, were cropped from the original footage. This duration was chosen to provide sufficient time for observers to evaluate multiple strides, which is essential for accurate lameness detection, while avoiding viewer fatigue or disengagement. The 20‐s video length was considered optimal for maintaining focus during repeated scoring tasks. The videos were duplicated three times to create triplicates for each animal. Each video was assigned a unique random number to blind the lecturer and observers to the lameness scores and identities of the cows. All videos were securely stored on a flash drive for subsequent use during the study.

2.3. Development of Lameness Scoring Forms

To facilitate scoring, forms were designed using Google Forms (https://forms.gle/SHMazUXMvoXbeRaF6), with each question corresponding to a specific video. Recruitment was conducted via public announcements on social media, through personal communication, and during routine university lessons. Interested senior veterinary students were instructed to contact the responsible researcher (YAO) for further details. Participants attended an informational briefing held at the university, during which study procedures and inclusion criteria (limited to senior veterinary students with no prior experience with bovine lameness) were explained. Participant lists were anonymised, and appointment slots for four observation sessions were filled with the enrolled students.

2.4. Survey Procedures

The survey link (https://forms.gle/SHMazUXMvoXbeRaF6) was distributed to participants. Prior to the start of the sessions, each student was required to complete an online informed consent form embedded in the Google Form link. The lameness scoring sessions were conducted in a controlled environment within the university's amphitheatre. A total of 201 senior veterinary students participated. All students were assembled in the amphitheatre and seated in rows, ensuring each participant had an unobstructed view of the projection screen. Video recordings of cows exhibiting various degrees of lameness were projected on a large screen using high‐resolution visual equipment, ensuring video quality and consistent viewing conditions. Each video clip, 20 s in length, was shown once at normal speed and without sound. The exclusion of audio was intentional to minimise potential bias caused by environmental noise, vocalisations, or handler cues that could influence the observer's perception. By eliminating auditory input, the study ensured that the scoring was based solely on visual assessment of the cow's locomotion, thereby enhancing the reliability and objectivity of the evaluation process. The videos were not replayed in slow motion to simulate real‐life clinical conditions where observers typically view cows at normal walking speed. After each video, students were given a fixed amount of time (approximately 30 s) to independently enter their lameness score using a Google Form accessed via their personal devices. The session was conducted in a single round, where students were shown a set of randomised video clips for lameness detection. Triplicate presentations of the same cows were included to evaluate intra‐rater reliability. No feedback was provided during the session to ensure independent evaluation, and participants were instructed not to communicate or discuss their observations with peers. Although all students viewed the same videos under identical conditions, their responses remained anonymous and unlinked, which ensured unbiased data collection. This approach allowed for efficient data collection under standardised conditions while preserving the integrity of individual assessments.

2.5. Statistical Analyses

Survey responses were extracted from Google Forms, and the number of responses was matched with the corresponding video triplicates. A 5‐point confusion matrix was constructed to compare students’ predicted lameness scores against the lameness score of experienced observers assigned to the cows. MedCalc statistical software (Version 20.015) was employed to calculate comprehensive performance metrics, including true positives (TP), false negatives (FN), false positives (FP), true negatives (TN), sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), prevalence, positive predictive value (PPV), negative predictive value (NPV), and accuracy for each lameness score (LS1–LS5). Overall accuracy was computed as the proportion of correct predictions across all categories. The data were transferred to SPSS 22 statistical software (SPSS Inc., Chicago, IL, USA) for agreement statistics, reliability analysis and binary logistic regression modelling. Intra‐rater reliability (the consistency of an individual student's scores across multiple assessments of the same cows) was assessed using the intraclass correlation coefficient (ICC). ICC values were calculated for each lameness score level (LS1–LS5) to evaluate the stability of individual scores across repeated observations. Inter‐rater reliability (the consistency of scores between different students) was assessed using Krippendorff's alpha, which was calculated for each lameness score. Binary logistic regression analysis was conducted to evaluate how lameness severity levels (LS1–LS5) affected students' detection accuracy by senior veterinary students. The outcome variable was coded as 1 for correct detection and 0 for incorrect detection. Prior to model fitting, key assumptions of binary logistic regression were evaluated. Multicollinearity among predictors was assessed using the variance inflation factor, with values below 5 indicating acceptable levels. Linearity of any continuous predictors with the logit of the outcome was checked using the Box–Tidwell test. The model's overall significance was evaluated using the likelihood ratio chi‐square test, and model fit was assessed with the Hosmer–Lemeshow goodness‐of‐fit test. The odds ratios (OR) and 95% confidence intervals (CI) were calculated to assess the likelihood of correct identification as lameness severity increased. All statistical tests were two‐tailed, and a significance level of p < 0.05 was considered statistically significant for all analyses.

3. Results

3.1. Study Participants

A total of 201 senior veterinary students participated in the study, representing 78.8% of the 255 eligible students from the academic years 2023–2024 and 2024–2025. The distribution of true and false scores based on lameness scores is illustrated in Figure 1.

FIGURE 1.

FIGURE 1

Open in a new tab

Distribution of true and false scores among 201 students and their 9045 responses across lameness scores.

3.2. Students’ Prediction Performance of Lameness Scores

The confusion matrix comparing students' lameness score predictions for each lameness score degree (LS1 to LS5) with the lameness score of experienced observers is presented in Table 1. The highest number of correct predictions occurred at LS5 (1372 correct scores), followed by LS1 (1347 correct scores). Conversely, the highest number of incorrect predictions was observed for LS3 (1126 incorrect scores) and LS4 (1105 incorrect scores).

TABLE 1.

Confusion matrix comparing students’ lameness score predictions to lameness score of experienced observer.

Lameness score of experienced observer n (%)
1 2 3 4 5 Total
Students’ predictions (%) 1 1347 (14.89) 580 (6.41) 216 (2.39) 9 (0.10) 6 (0.07) 2158 (23.86)
2 430 (4.75) 1102 (12.18) 725 (8.01) 157 (1.74) — 2414 (26.69)
3 15 (0.17) 110 (1.22) 683 (7.55) 846 (9.35) 15 (0.17) 1669 (18.45)
4 4 (0.04) 5 (0.05) 167 (1.85) 704 (7.78) 416 (4.60) 1296 (14.33)
5 13 (0.14) 12 (0.13) 18 (0.20) 93 (1.03) 1372 (15.17) 1508 (16.67)
Total n (%) 1809 (20) 1809 (20) 1809 (20) 1809 (20) 1809 (20) 9045 (100)
Open in a new tab

Performance metrics for students’ lameness score predictions are shown in Table 2. Students achieved the highest sensitivity and accuracy for lameness scores of 1 (sensitivity: 74.46%; accuracy: 85.93%) and 5 (sensitivity: 75.84%; accuracy: 93.67%). However, predictions for LS3 exhibited the lowest sensitivity (37.76%; 95% CI: 35.52–40.04) and accuracy (76.65%). The highest specificity was recorded for LS5 (98.12%), whereas LS2 showed the lowest specificity (81.87%).

TABLE 2.

Performance metrics for students’ lameness score predictions.

LS TP FN FP TN Sensitivity % (95% CI) Specificity % (95% CI) PLR NLR Prevalence % (95% CI) PPV % NPV % Accuracy % (95% CI)
Not lame 1 1347 462 811 6425 74.46 (72.39‐76.46) 88.79 (88.04‐89.51) 6.64 0.29 20 (19.18‐20.84) 62.42 93.29 85.93 (85.19‐86.64)
Lame 2 1102 707 1312 5924 60.92 (58.63‐63.17) 81.87 (80.96‐82.75) 3.36 0.48 20 (19.18‐20.84) 45.65 89.34 77.68 (76.81‐78.53)
3 683 1126 986 6250 37.76 (35.52‐40.04) 86.37 (85.56‐87.16) 2.77 0.72 20 (19.18‐20.84) 40.92 84.73 76.65 (75.76‐77.52)
4 704 1105 592 6644 38.92 (36.66‐41.21) 91.82 (91.16‐92.44) 4.76 0.67 20 (19.18‐20.84) 54.32 85.74 81.24 (80.42‐82.04)
5 1372 437 136 7100 75.84 (73.80‐77.80) 98.12 (97.78‐98.42) 40.35 0.25 20 (19.18‐20.84) 90.98 94.20 93.67 (93.14‐94.16)
Open in a new tab

Abbreviations: CI, confidence interval; FN: false negative; FP: false positive; LS: lameness score; NLR: negative likelihood ratio; NPV: negative predicted value; PLR: positive likelihood ratio; PPV: positive predicted value; TN: true negative; TP: true positive.

3.3. Students’ Inter‐Rater Reliability in Lameness Scores

The agreement and inter‐rater reliability of senior veterinary students’ lameness scoring were evaluated across different degrees of lameness (Table 3). Exact agreement with the lameness score of experienced observers varied by lameness degree, ranging from 37.76% to 76.67%. The highest agreement was observed for LS1 (not lame) at 76.67% and LS5 (severe lame) at 75.84%, while lower agreement was noted for LS3 and LS4 (37.76% and 38.92%, respectively). Krippendorff's alpha coefficient ranged from 0.106 (score 5) to 0.490 (score 3), indicating generally low to moderate reliability. Mean scores assigned by students for each lameness score degree are presented in Table 3.

TABLE 3.

Agreement and reliability metrics for students' lameness score Predictions across various lameness scores.

LS Exact agreement with lameness score of experienced observer (%) Krippendorff's alpha coefficient Mean score (SD)
Not lame 1 76.67 0.371 1.29 (0.57)
Lame 2 60.92 0.315 1.77 (0.63)
3 37.76 0.490 2.47 (0.86)
4 38.92 0.362 3.40 (0.74)
5 75.84 0.106 4.74 (0.50)
Open in a new tab

Note: Krippendorff's alpha coefficient values ≥ 0.80 indicate satisfactory agreement, 0.67–0.79 moderate agreement, and < 0.67 poor agreement.

Abbreviations: LS, lameness score; SD, standard deviation.

3.4. Students’ Intra‐Rater Reliability in Lameness Scores

Intra‐rater reliability for repeated assessments of the same animal revealed the highest intraclass correlation coefficient (ICC) for LS4 (0.798), followed by LS3 (0.775). In contrast, LS5 (0.613) and LS2 (0.625) demonstrated the lowest ICC values. (Table 4).

TABLE 4.

Intra‐Rater reliability metrics for students' lameness score predictions across lameness scores.

LS Observation number Mean score (SD) ICC (95% CI)
Not lame 1 1 1.28 (0.54) 0.679 (0.364–0.462)
2 1.28 (0.59)
3 1.31 (0.57)
Lame 2 1 1.85 (0.69) 0.625 (0.569–0.674)
2 1.69 (0.61)
3 1.76 (0.58)
3 1 2.63 (0.82) 0.775 (0.736–0.808)
2 2.41 (0.88)
3 2.37 (0.85)
4 1 3.49 (0.74) 0.798 (0.765–0.827)
2 3.28 (0.75)
3 3.41 (0.71)
5 1 4.71 (0.54) 0.613 (0.556–0.663)
2 4.75 (0.47)
3 4.76 (0.49)
Open in a new tab

Note: ICC values < 0.5 indicate poor, 0.5–0.75 moderate, 0.75–0.9 good, and > 0.9 excellent reliability.

Abbreviations: ICC, intraclass correlation coefficient; LS, lameness score; SD, standard deviation.

3.5. Logistic Regression Analysis of Students’ Accuracy Predicting Lameness Identification

Logistic regression analysis revealed significant differences in students' accuracy when predicting lameness based on lameness scores (Table 5). A one‐unit increase in the lameness score of the experienced observer was associated with a decrease in student accuracy, with an odds ratio of 0.924 (95% CI: 0.897–0.952).

TABLE 5.

Logistic regression analysis of the effect of lameness scores on student accuracy.

Variables b SE Wald p‐value OR (95% CI)
Overall LS −0.079 0.015 27.359 < 0.001 0.924 (0.897‐0.952)
LS2 vs. LS1 −0.074 0.077 0.925 0.336 0.929 (0.799‐1.080)
LS3 vs. LS1 −0.700 0.073 91.839 < 0.001 0.496 (0.430‐0.573)
LS4 vs. LS1 −1.644 0.073 503.372 < 0.001 0.193 (0.167‐0.223)
LS5 vs. LS1 −1.595 0.073 476.124 < 0.001 0.203 (0.176‐0.234)
Open in a new tab

Abbreviations: b, estimated coefficient; CI, confidence interval; OR, odds ratio; LS, lameness score; SE, standard error.

When comparing lameness scores to not lame cows (LS1), the odds of correct identification for LS3 were reduced to 49.6% (p < 0.001). For LS4 and LS5, the odds further decreased to 19.3% and 20.3%, respectively (p < 0.001 for both). However, no significant difference was observed for LS2, with an odds ratio of 0.929 (p = 0.336).

4. Discussion

Early detection of lameness is paramount in dairy farming due to its association with prolonged healing times, increased treatment costs, and significant losses in productivity and reproductive performance. (Huxley 2013; Thomas et al. 2016; Dolecheck and Bewley 2018; Akin and Akin 2018) Despite advancements in automated lameness detection systems, their adoption remains limited due to high installation costs and operational complexities. (Flower and Weary 2009; Alsaaod et al. 2019; Kang et al. 2021; Akin et al. 2024) Consequently, visual lameness detection continues to be the gold standard. However, this method is susceptible to observer variability and experience‐dependent accuracy. (Sprecher et al. 1997; O'Callaghan et al. 2003; Espejo et al. 2006; Flower and Weary 2006; Lambertz 2019) The current study underscores the importance of equipping senior veterinary students with the skills necessary for accurate lameness detection under practical field conditions. Given that delayed detection of mild lameness can lead to more severe outcomes, including increased treatment costs and culling, it is critical that veterinary curricula address these challenges effectively. (; Leach et al. 2013; Renn et al. 2014; Kang et al. 2021)

Combining theoretical teaching methods with methodologies that incorporate visual outputs, such as video‐assisted learning, has demonstrated promise in enhancing students' gait assessment skills in various species, including humans, horses, pigs, and dogs. (Keegan et al. 1998; Main et al. 2000; Read et al. 2003; Brunnekreef et al. 2005; Dahmer 2007; Williams et al. 2009) In the present study, video‐assisted learning was an integral component of the teaching curriculum, aligning with the educational standards of EAEVE and VEDEK, which are monitored and applied globally in veterinary faculties. This study aimed to evaluate the reliability and accuracy of senior veterinary students’ bovine lameness gait assessment skills as part of their diagnostic training, having completed their required pre‐graduation education. Notably, while earlier research has primarily focused on experienced veterinarians (Brenninkmeyer et al. 2007; Thomsen et al. 2008; Tunstall et al. 2021), this study addresses a gap by evaluating the effectiveness of lameness detection training in veterinary students. Our findings highlight the importance of incorporating innovative teaching tools and methodologies, such as increased hands‐on training, into veterinary curricula. These findings provide a basis for future studies aimed at optimising veterinary training programmes to enhance early lameness detection and ultimately improve animal welfare and farm outcomes.

Senior veterinary students exhibited higher accuracy in detecting healthy cows (LS1) and severe lameness (LS5) cases compared to intermediate lame cases (LS2 to LS4), with accuracies of 85.93% and 93.67% and sensitivities of 74.46% and 75.84%, respectively (Table 2). Sensitivity for LS3 was particularly low at 37.76%, and the accuracy for LS3 and LS4 declined to 76.65% and 78.45%, respectively. These differences may arise from the subtler clinical signs of intermediate lameness cases, which require more refined observational skills to identify. This finding aligned with previous studies that even experienced professionals may struggle with detecting intermediate lameness cases. (Main et al. 2000; Brunnekreef et al. 2005; Thomsen et al. 2008; Flower and Weary 2009; Leach et al. 2013). Also, students exhibited higher specificity than sensitivity across all lameness scores (Table 2), indicating students eager to have a conservative diagnostic approach that minimises false positives but increases the likelihood of underdiagnosis. This cautious approach, while reducing unnecessary interventions, risks delaying appropriate treatment and may compromise animal welfare. The highest accuracy observed for LS5 reflects the clear distinction in movement patterns for cows with severe lameness, consistent with earlier observations that lameness severity influences detection accuracy. (Sprecher et al. 1997; Flower and Weary 2006)

The present results reveal notable variability in inter‐ and intra‐rater agreement among students when scoring lameness (Tables 3 and 4). The highest inter‐rater reliability, as measured by Krippendorff's alpha, was observed for LS3 (moderate lameness), despite this score having the lowest exact agreement with the lameness score of the experienced observer. Similarly, LS4 showed moderate reliability but low accuracy. These findings suggest that although students were relatively consistent among themselves when evaluating moderate lameness, their scores often deviated from the lameness scores of experienced observers. In contrast, LS1 (not lame) and LS5 (severe lame) demonstrated high levels of agreement with lameness score of experienced observer but low reliability, implying that while these scores were easier to recognise, consistency in interpretation was limited. The intra‐rater agreement was also highest for LS3 and LS4 but declined for LS2 and LS5, indicating that individual scoring consistency was strongest at mid‐range severity levels. These outcomes suggest that current teaching methods have partially succeeded in delivering a uniform framework for lameness scoring. However, discrepancies between agreement and reliability underscore the need for more targeted training. Logistic regression analysis further showed that students’ accuracy in identifying lameness decreased as the severity increased (Table 5). This was particularly evident in LS4 and LS5, where students frequently underestimated the lameness severity, and in LS2, where students struggled to differentiate between sound and mildly lame cows. These challenges align with prior research indicating that inexperienced observers perform better in broad classifications than in nuanced scoring systems. (Thomsen et al., 2008) Such difficulty may stem from educational gaps that do not sufficiently prepare students to assess subtle gradations in lameness, a limitation that could persist into clinical practice. To address these gaps, future veterinary curricula should provide enhanced opportunities for students to refine their lameness detection skills. Strategies such as expanded video‐based training, hands‐on scoring in farm environments, regular farm visits, and structured expert feedback sessions are recommended. These approaches aim to improve diagnostic accuracy and inter‐observer consistency, ultimately promoting better animal welfare and more informed herd management.

This study has several limitations that should be addressed in future research. First, the use of video recordings instead of live observations may have influenced students' assessments, as videos may not capture the full range of behavioural context observed during live evaluations. Additionally, the study was conducted at a single institution, which may limit the generalisability of the findings to other veterinary schools or regions with different teaching methodologies. Furthermore, the reliance on a single experienced observer to assign the “true” lameness scores introduces the potential for observer bias, which could have influenced the comparison of students' performance metrics. Including multiple experienced observers and comparing their scores could provide a more robust measure of accuracy. Future studies should also examine the effectiveness of video‐based training in improving the detection of milder lameness cases, as well as the impact of different teaching strategies—such as repeated practice, expert feedback, or simulated clinical scenarios—on enhancing diagnostic accuracy and inter‐observer consistency. Longitudinal studies tracking students' progress over the course of their education, alongside cross‐institutional studies comparing training outcomes, would offer valuable insights into how lameness detection skills develop over time and across varying educational settings. Another important limitation of this study is the reliance on the Sprecher 5‐point scale to assess lameness, which may have influenced the results. While this scale is widely used and relatively simple, it may not capture the full spectrum of lameness severity, particularly in cases of mild or subtle gait abnormalities. Moreover, the use of a discrete ordinal scale rather than a continuous measure could limit the sensitivity of the scoring system, potentially leading to inconsistencies in how different observers classify cows with similar gait deviations. Future research could benefit from incorporating multiple scoring systems and exploring their comparative reliability and sensitivity in detecting lameness, helping to better understand the limitations inherent in any single scoring method.

5. Conclusion

This study highlights the importance of developing reliable and accurate methods for lameness detection in dairy cows, particularly in veterinary education. It shows that senior veterinary students can achieve reasonable accuracy in detecting lameness, though performance varies with lameness severity. The findings emphasise the need for ongoing advancements in educational strategies to enhance perceptual accuracy and clinical judgement, with a focus on video‐assisted tools. Given the challenges in detecting cases across varying lameness severities, future training programmes should incorporate more comprehensive visual and practical learning experiences to improve diagnostic skills. Persistent difficulties in scoring intermediate lameness cases underscore the need to refine teaching methods further.

Author Contributions

Yalcin Alper Ozturan: conceptualisation, investigation, funding acquisition, writing – original draft, methodology, validation, visualisation, writing – review and editing, software, formal analysis, project administration, data curation, supervision and resources. Ibrahim Akin: conceptualisation, investigation, writing – review and editing, methodology, validation, formal analysis and supervision.

Ethics Statement

This study was approved by the Ethical Committee of Aydin Adnan Menderes University (Approval No: 64583101/2024/103).

Conflicts of Interest

The authors declare no conflicts of interest.

Funding: The authors received no specific funding for this study.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author (Yalcin Alper OZTURAN), upon reasonable request.

References

  1. Akin, I. , and Akin T.. 2018. “Economic Impact of Digital Dermatitis Treatment on a Dairy Farm: An Application of the Break‐Even Analysis.” Ciencia Rural 48, no. 8: e2017079. [Google Scholar]
  2. Akin, I. , Kalkan Y., and Ozturan Y. A.. 2024. “Vertical Back Movement of Cows During Locomotion: Detecting Lameness With a Simple Image Processing Technique.” Journal of Dairy Research 931, no. 3: 278–285. [DOI] [PubMed] [Google Scholar]
  3. Alsaaod, M. , Fadul M., and Steiner A.. 2019. “Automatic Lameness Detection in Cattle.” Veterinary Journal 246: 35–44. [DOI] [PubMed] [Google Scholar]
  4. Barstow, A. , Pfau T., Bolt D. M., Smith R. K., and Weller R.. 2014. “Design and Validation of a Computer‐Aided Learning Program to Enhance Students' Ability to Recognize Lameness in the Horse.” Journal of Veterinary Medical Education 41, no. 1: 1–8. [DOI] [PubMed] [Google Scholar]
  5. Bell, N. 2015. “The Farm Audit: Foot Health, Lameness and Footcare.” In Bovine Medicine, edited by Cockcroft P. D., John Wiley & Sons, 406–417. [Google Scholar]
  6. Brenninkmeyer, C. , Dippel S., March S., Brinkmann J., Winckler C., and Knierim U.. 2007. “Reliability of a Subjective Lameness Scoring System for Dairy Cows.” Animal Welfare 16, no. 2: 127–129. [Google Scholar]
  7. Brunnekreef, J. J. , van Uden C. J., van Moorsel S., and Kooloos J. G. M.. 2005. “Reliability of Videotaped Observational Gait Analysis in Patients With Orthopedic Impairments.” BMC Musculoskeletal Disorders [Electronic Resource] 6: 17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Dahmer, J. 2007. “Lernen.” In Didaktik der Medizin Professionelles Lehren fördert effektives Lernen, edited by Dahmer J., Schattauer Verlag, 1–9. [Google Scholar]
  9. Dolecheck, K. , and Bewley J.. 2018. “Animal Board Invited Review: Dairy Cow Lameness Expenditures, Losses and Total Cost.” Animal 12, no. 7: 1462–1474. [DOI] [PubMed] [Google Scholar]
  10. Espejo, L. A. , Endres M. I., and Salfer J. A.. 2006. “Prevalence of Lameness in High‐producing Holstein Cows Housed in Freestall Barns in Minnesota.” Journal of Dairy Science 89, no. 8: 3052–3058. [DOI] [PubMed] [Google Scholar]
  11. Flower, F. C. , and Weary D. M.. 2006. “Effect of Hoof Pathologies on Subjective Assessments of Dairy Cow Gait.” Journal of Dairy Science 89, no. 1: 139–146. [DOI] [PubMed] [Google Scholar]
  12. Flower, F. C. , and Weary D. M.. 2009. “Gait Assessment in Dairy Cattle.” Animal 3, no. 1: 87–95. [DOI] [PubMed] [Google Scholar]
  13. Geber, P. , and Wilson M.. 2010. “Teaching at the Interface of Dance Science and Somatics.” Journal of Dance Medicine & Science 14, no. 2: 50–57. [PubMed] [Google Scholar]
  14. Huxley, J. N. 2013. “Impact of Lameness and Claw Lesions in Cows on Health and Production.” Livestock Science 156, no. 1‐3: 64–70. [Google Scholar]
  15. Kang, X. , Zhang X. D., and Liu G. A.. 2021. “Development of Computer Vision‐Based Lameness Detection for Dairy Cows and Discussion of the Practical Applications .” Sensors 21, no. 3: 753. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Kawamura, C. M. , de Morais Filho M. C., Barreto M. M., Asa S. K. P., and Novo N. F.. 2007. “Comparison Between Visual and Three‐Dimensional Gait Analysis in Patients With Spastic Diplegic Cerebral Palsy.” Gait & Posture 25, no. 1: 18–24. [DOI] [PubMed] [Google Scholar]
  17. Keegan, K. G. , Wilson D. A., Wilson D. J., et al. 1998. “Evaluation of Mild Lameness in Horses Trotting on a Treadmill by Clinicians and Interns or Residents and Correlation of Their Assessments With Kinematic Gait Analysis.” American Journal of Veterinary Research 59, no. 11: 1370–1377. [PubMed] [Google Scholar]
  18. Lambertz, C. 2019. “Using Cluster Analysis to Differentiate Degrees of Lameness in Cattle.” The Veterinary Record 184, no. 7: 218. [DOI] [PubMed] [Google Scholar]
  19. Leach, K. A. , Paul E. S., Whay H. R., et al. 2013. “Reducing Lameness in Dairy Herds—Overcoming Some Barriers.” Research in Veterinary Science 94, no. 3: 820–825. [DOI] [PubMed] [Google Scholar]
  20. Main, D. C. , Clegg J., Spatz A., and Green L. E.. 2000. “Repeatability of a Lameness Scoring System for Finishing Pigs.” The Veterinary Record 147, no. 20: 574–576. [DOI] [PubMed] [Google Scholar]
  21. O'Callaghan, K. A. , Cripps P. J., Downham D. Y., and Murray R. D.. 2003. “Subjective and Objective Assessment of Pain and Discomfort Due to Lameness in Dairy Cattle.” Animal Welfare 12, no. 4: 605–610. [Google Scholar]
  22. Read, H. S. , Hazlewood M. E., Hillman S. J., Prescott R. J., and Robb J. E.. 2003. “Edinburgh Visual Gait Score for Use in Cerebral Palsy.” Journal of Pediatric Orthopedics 23, no. 3: 296–301. [PubMed] [Google Scholar]
  23. Renn, N. , Onyango J., and Mccormick W.. 2014. “Digital Infrared Thermal Imaging and Manual Lameness Scoring as a Means of Lameness Detection in Cattle.” Veterinary Clinical Sciences 2, no. 2: 16–23. [Google Scholar]
  24. Schieder, K. , Zsoldos R. R., Dippel M., Siedler C., Tichy A., and Licka T. F.. 2020. “Use of Physical Self‐Experience for Teaching Lameness Evaluation: Short‐Term Effects on Lameness Evaluation of Horses With Mild Forelimb Lameness by Novice Veterinary Students.” Journal of Veterinary Medical Education 47, no. 3: 342–355. [DOI] [PubMed] [Google Scholar]
  25. Sprecher, D. J. , Hostetler D. E., and Kaneene J. B.. 1997. “A Lameness Scoring System that Uses Posture and Gait to Predict Dairy Cattle Reproductive Performance.” Theriogenology 47, no. 6: 1179–1187. [DOI] [PubMed] [Google Scholar]
  26. Starke, S. D. , and May S. A.. 2017. “Veterinary Student Competence in Equine Lameness Recognition and Assessment: A Mixed Methods Study.” The Veterinary Record 181, no. 7: 168–175. [DOI] [PubMed] [Google Scholar]
  27. Thomas, H. J. , Remnant J. G., Bollard N. J., et al. 2016. “Recovery of Chronically Lame Dairy Cows Following Treatment for Claw Horn Lesions: A Randomised Controlled Trial.” The Veterinary Record 178, no. 5: 116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Thomsen, P. T. , Munksgaard L., and Tøgersen F. A.. 2008. “Evaluation of a Lameness Scoring System for Dairy Cows.” Journal of Dairy Science 91, no. 1: 119–126. [DOI] [PubMed] [Google Scholar]
  29. Tunstall, J. , Mueller K., Grove‐White D., Oultram J. W. H., and Higgins H. M.. 2021. “Lameness in Beef Cattle: A Cross‐Sectional Descriptive Survey of On‐Farm Practices and Approaches.” Frontiers in Veterinary Science 8: 657299. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Waxman, A. S. , Robinson D. A., Evans R. B., Hulse D. A., Innes J. F., and Conzemius M. G.. 2008. “Relationship Between Objective and Subjective Assessment of Limb Function in Normal Dogs With an Experimentally Induced Lameness.” Veterinary Surgery 37, no. 3: 241–246. [DOI] [PubMed] [Google Scholar]
  31. Weaver, A. D. , Atkinson O., St Jean J., and Steiner A.. 2018. “Musculoskeletal Conditions and Lameness.” In Bovine Surgery and Lameness, edited by Weaver A. D., Atkinson O., St Jean G., and Steiner A., John Wiley & Sons. 267–350. [Google Scholar]
  32. Williams, G. , Morris M. E., Schache A., and McCrory P.. 2009. “Observational Gait Analysis in Traumatic Brain Injury: Accuracy of Clinical Judgment.” Gait & Posture 29, no. 3: 454–459. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author (Yalcin Alper OZTURAN), upon reasonable request.

Articles from Veterinary Medicine and Science are provided here courtesy of Wiley

RESOURCES