Sensitivity and specificity of mobility scoring for the detection of foot lesions in pasture based Irish dairy cows

Lameness is an important production disease in dairy cows worldwide and has detrimental effects on cows’ welfare, production, and reproductive performance, thus impacting the sustainability of dairy farming. Timely and effective detection of lameness allows for effective treatment, minimizing progression of disease and maximizing the prognosis of recovery. Mobility Scoring (MS) is a 4 point (0–3) visual lameness scoring system that is the industry standard in several countries. However, few studies have examined the sensitivity (Se) and specificity (Sp) of MS to detect foot lesions. The aim of this observational study was to determine the Se and Sp of MS to detect foot lesions in dairy cattle in a pasture-based system. 595 primi-and multiparous cows were randomly selected from 12 commercial Irish dairy farms and recruited for the study. Recruited cows were mobility scored and passed through a foot paring crate where all 4 feet were lifted for examination. The team recorded the anatomical location and severity of any foot lesions present based on appearance only. Then, based on the type and severity of the lesions present, cows were classified according to 3 case definitions Case Definition 1: Any lesion present; Case Definition 2: Moderate lesions present (excluding minor lesions expected to have a low probability of impacting gait); and Case Definition 3: Severe lesions present (only including lesions most likely to result in a detectable gait abnormality). Sensitivity and specificity of MS was calculated based on a threshold of MS ≥2, defined as impaired (MS = 2) or severely impaired (MS = 3) mobility for each of the 3 case definitions, at the overall level and disaggregated by parity. The overall cow-level lesion prevalence based on the Case Definition 1 was 0.54 with significant between-herd variation. The overall Se and Sp of MS for the detection of foot lesions, was 0.18 and 0.96; 0.35 and 0.94; 0.43 and 0.94 for the in Case Definitions 1, 2 and 3 respectively. Our findings showed poor sensitivity, but high specificity of MS for the detection of cows with foot lesions in a pasture-based system.


INTRODUCTION
Lameness is described as the inability to express a physiological locomotion pattern in one or more limbs, most frequently as a consequence of pain (Oehm et al., 2019).Lameness is a significant welfare issue and economic burden on dairy farms.It can significantly compromise the "five freedoms" of animal welfare and negatively impacts the time budget of cows (Whay and Shearer., 2017).The economic impact of lameness has been well documented in an Irish pasture-based system (Ryan and O'Grady, 2004;O'Connor et al., 2020b).The cost associated with lameness arises from its effects on production and fertility (Warnick et al., 2001;O'Connor et al., 2020b), costs of treatment, and costs of time and labor in addressing lameness by stock personnel (O'Connor et al., 2023;Robcis et al., 2023;Bruijnis et al., 2010).
Early detection of lameness enables prompt treatment, and the initiation of preventive measures and is an important aspect of safeguarding animal welfare on dairy farms both in housed and in pasture-based cows (Whay and Shearer, 2017).Effective detection of lameness at an early stage is important to prevent progression to severe lameness that can reduce the prognosis of treatment.Prompt identification of cases and appropriate intervention can help minimize direct and indirect costs associated with lameness (Van Nuffel et al., 2015).Routine lameness prevalence monitoring is advised and recommended as a component of animal welfare auditing in some dairy farm assurance schemes such as the Red Tractor Farm UK national dairy assurance scheme (Mullan et al., 2021).
Several methods can be used to detect lameness in cattle.Automated detection systems based on gait analysis and computer vision (Kang et al., 2020), weight distribution (Pastell et al., 2010), accelerometry (O'Leary et al., 2020) and thermal imaging (Werema et al., 2021), have been developed with varying levels of success.The development of automated lameness detection methods is an ongoing area of research.
At present however, observer-based detection methods are most commonly used in the industry.These methods use assessment criteria to interpret visual observation of gait according to a series of score criteria or descriptions relating to gait abnormality severity.A range of broadly similar scoring systems are used.For example, the Sprecher locomotion scoring system is a 5-point scale (1-5) (Sprecher et al., 1997) and includes a requirement to assess back posture while standing, whereas the Agriculture and Horticulture Development Board, UK, (AHDB) mobility scoring system is a 4-point scale (0-3) and does not require an assessment of the standing animal.The AHDB mobility scoring (MS) system has been adopted as the UK dairy industry standard mobility scoring system (Afonso et al., 2020), and has gained popularity in Ireland (O'Connor et al., 2020a;Browne et al., 2022c) and other dairy producing countries.Importantly, this system overcomes the need to assess the back posture of cows at a stance, which requires an additional assistant to stop cows and isn't always convenient in a pasture-based system where most lameness scoring is carried out as cows are walking back to pasture after milking (Fabian et al., 2014).
Several studies have evaluated visual lameness scoring methods in terms of both inter-and intra-observer agreement and have shown varying results.Gardenier et al. (2021) found that the Australian Healthy Hooves 4-level locomotion scoring system, which is similar to the AHDB Mobility score, has an inter-and intraobserver percentage agreement of 79% and 82% respectively when using relative pairwise scoring.Garcia et al. (2015) have shown varying levels of intra-operator repeatability, using a 5-point scale based on a revised version of the Sprecher (1997) locomotion score.Schlageter-Tello et al. (2014) found that acceptable inter-rater reliability of ≥ 75% using the Flower and Weary (2006) 5-point visual lameness scoring system was attained only when it was condensed into 2 categories of 'lame' and 'non-lame'.
While a significant body of literature has developed on the basis of the AHDB Mobility scoring system, there have been limited studies investigating the sensitivity and specificity of MS, or other visual gait assessment methods, for the detection of foot lesions in dairy cattle.Foot lesions are indicative of pathology which is traditionally the focus of treatment decisions.Although the presence of foot lesions may not be sufficient to cause locomotion disruption, a poor sensitivity would indicate that significant numbers of cattle have foot lesions that may be painful, and/or benefit from treatment, and yet are not detected using the industry standard method for lameness detection, with the potential for significant welfare implications.On the other hand, a poor specificity of mobility scoring for the detection of foot lesions, would lead to many animals being treated, or inspected for treatment in the absence of a treatable, or explainable, lameness-causing foot lesion.Finally, at a herd level, if mobility scores are used to assess welfare on farms, the potential for a mismatch between the herd-level prevalence of foot lesions and aggregated mobility scores has implications for the utility of this scoring system for herd-level welfare assessment.
Therefore, the objective of this observational study was to determine the sensitivity and specificity of mobility scoring for the detection of foot lesions on commercial pasture-based dairy farms.

Farm recruitment
Twelve spring-calving pasture-based dairy herds in Ireland were recruited to take part in this trial during the summer months of 2021.Six of these herds were in county Meath, 3 in Kildare, one in Offaly, one in Wicklow and one in Cavan.One farm used a robotic milking system, with the rest of the farms using a conventional manual parlor with cows milked twice daily.Farm recruitment was based on a convenience sample of herds willing to participate in the research study.

Cow level data collection
Cow identification lists for each herd were extracted from the Irish Cattle Breeding Federation (ICBF) database, and a randomized sampling list of 50 cows per herd was generated with the use of a Microsoft Office Excel random number generator (Microsoft Excel for Microsoft 365, Version 2302).
The sampling procedure on each farm consisted of 2 separate visits, 2 weeks apart.During the initial visit, the 50 randomly chosen cows were identified and an accelerometer fitted to the hindlimb.This step was for the purpose of another study, and 50 cows were chosen per herd as this was the total number of accelerometers available for that study.During the second visit to the farm 2 weeks later; the recruited cows were drafted out from the rest of the herd following milking.On 7 farms this was carried out with the use of automatic drafting and on the 5 other farms the cows were manually separated as they emerged from the milking parlor.Fol-lowing separation of the sample cohort, each cow in the trial was visually assessed and assigned a MS using the AHDB Dairy mobility scoring system (AHDB, 2020).

Foot inspection
An experienced hoof paring professional was recruited for the trial provided by Farm Relief Services, (Ballyjamesduff, Co. Cavan, Ireland) who attended each second day visit, i.e., on the day on which MS was conducted.MS was conducted after milking and immediately before the cows being drafted for foot examination.Each of the recruited cows passed through a hydraulic foot-paring crate where each cow's feet were lifted sequentially in the standing position.Each of the 4 feet were lifted, cleaned, and examined for the presence or absence of foot lesions.Once a lesion was detected, it was identified, graded and noted by the first author (a veterinarian and European College of Bovine Health Management resident).The ICAR claw health atlas (Egger-Danner et al., 2015) was reviewed and discussed by the first author and the study co-authors before study commencement but was not carried to farm visits, instead the lesion category descriptions detailed below were used as the only aid for identification on farm.Previous lameness history was not available and therefore not considered in the analysis.

Lesion grading
For white line disease, solar hemorrhage and claw overgrowth, severity grading scales were developed and agreed upon with the co-authors.These scales were developed based on existing severity scales from the literature and modified according to the clinical experience of the study team in foot lesion presentations on commercial dairy farms.The full description of the grading scales used are provided as supplementary material.
White line disease, as described by Shearer et al., (2017), was recorded on a severity scale of 1 -3).Grade 1 corresponded with obvious blackening of the white line with small patches or specks of dirt; grade 2 was when one or multiple larger areas of blackening occur within the white line; and grade 3 corresponded to areas of blackening with dirt within the white line including the emergence of abscessation proximally at the coronary band or communication from the white line lesion to a double (under-run) sole.
Solar hemorrhage was recorded based on Greenough and Vermunt (1991), with modification to grade 4: grade 1 was described as yellow discoloration; grade 2 as moderate red or pink discoloration; grade 3 described as severe hemorrhage; and grade 4 as severe hemorrhage with an associated soft spot but not a full thickness defect in the corium.
Heel erosion / slurry heel was also recorded using a 1 -4 severity scale with grades 1 and 2 defined as multiple shallow irregular depressions and multiple deep irregular depressions of the heel bulbs respectively.Grade 3 was described as shallow oblique grooves and grade 4 being deep oblique grooves with complete loss of structure of the heel (Smilie et al., 1999).
Claw overgrowth was recorded as a lesion within this study and encompassed excessive dorsal hoof wall length more than 8cm and the angle of the dorsal wall being < 50°.However, it also included axial groove overgrowth resulting in inappropriate weight bearing conformation of the digit.This lesion was recorded on a scale of 1-3.Grade 1 corresponded to toe length of 8 -10cm and no axial overgrowth; grade 2 corresponded with toe length of 8-10cm with some axial overgrowth; and grade 3 being toe length in excess of 10cm with severe axial overgrowth.
Digital dermatitis was recorded using the M-Stage lesion grading scale as described by Dopfer et al. (1997).
For the remaining lesions, solar ulcer, interdigital dermatitis, interdigital necrobacillosis, interdigital necrobacillosis, toe necrosis, trauma and double sole, no grading system was used, instead only the presence or absence was recorded.The decision not to use severity scales for these lesions was based on a combination of the absence of an established scoring system and whether differences in the severity in these lesion types would be expected to result in a change in case definition as outlined below.Trauma was also recorded as a lesion and consisted of any physical trauma to a digit not encompassed by any of the aforementioned lesions.Vertical wall fissures, axial wall fissures, horizontal wall fissures, broken hoof wall horn, penetrative sole injuries and loose hoof wall capsule were all recorded under trauma as simply present or absent.

Recording on farm
Foot lesions were recorded independently for each claw, on each of the 4 feet.The foot lesion and associated severity grade if applicable, and mobility score of each cow was recorded using a bespoke Filemaker Pro package (Filemaker Pro Advanced, Claris International Inc.) on a hand-held computer.The data from each farm was subsequently extracted as a Microsoft Office Excel file with each lameness lesion type corresponding to a column of data.The severity grade of each lesion where applicable, was recorded in the lesion column.

Case definition
A wide range of foot lesions may be present in dairy cows.It is understood that some foot lesions are expected to result in greater gait disturbance than others; for example, solar ulcers have been shown to be associated with greater behavioral signs of lameness compared with digital dermatitis or solar hemorrhage (Jewell et al., 2021).It was therefore expected that the ability of MS to detect foot lesions may vary according to the type and severity of the foot lesion present.Therefore, cows with lesions present on their feet were classified into 3 case definitions: Case Definition 1: Any lesion present; Case Definition 2: Moderate lesions present (excluding minor lesions expected to have a low probability of impacting gait); and Case Definition 3: Severe lesions present (only including lesions most likely to result in a detectable gait abnormality).Lesions type and severity grades were mapped to these case definitions according by a consensus agreement of the co-authors that were diplomates of the European College of Bovine Health Management (CIMA; CGMA; ER; LOG) (Table 2).Cases were identified for each of the 3 case definitions based on whether cows had one or more lesions that met the thresholds shown in Table 2.In the case of multiple lesions being identified in the same cow, the cow was assigned the category associated with the highest scoring lesion.For example, a cow having an M3 digital dermatitis lesion (considered to fulfil the criteria for Case definition 1, but not 2 or 3) and a solar ulcer (considered to fulfil the criteria for all 3 case definitions), was categorized as having fulfilled the criteria for all 3 case definitions.

Calculation of sensitivity and specificity
Descriptive statistics were carried out in Microsoft Excel.Prevalence of any lesion was calculated at the overall, parity and farm levels.For the purpose of evaluating the characteristics of MS as a diagnostic test, a 'true positive' was defined as an animal meeting or exceeding the criteria of the relevant case definition, and having a MS ≥2; a 'false positive' was defined as an animal not meeting the criteria of the relevant case definition yet having a MS ≥2; a 'true negative' was defined as an animal not criteria of the relevant case definition, and having a MS <2; while a 'false negative' was an animal meeting or exceeding the criteria of the relevant case definition, yet having a MS <2.
Sensitivity and specificity were calculated based on the proportion of cows for each of the case definitions that had a MS ≥2.These values were calculated at the overall level as well as broken down into cohorts by parity.95% Confidence interval (95% CI) for these values were calculated using the Clopper-Pearson (exact) method for binomial confidence interval.Positive and negative predictive values were calculated as the probability that a cow with a MS ≥2 had a foot lesion corresponding to the relevant case definition, and the probability that cow with a MS <2 did not have a foot lesion corresponding to the relevant case definition respectively.Confidence Intervals for predictive values were calculated using the method proposed by Mercaldo et al. (2007).Data manipulation was performed in Excel and R (R Core Team, 2019), summary statistics and calculation of confidence intervals was performed in R using the 'binom' package (Dorai-Raj, 2014), and the 'bdpv' package (Schaarschmidt, 2019).

Descriptive statistics
The final study population was 595 cows in total from 12 farms.Herd size ranged from 91 to 490 cows, with a mean of 243 cows.The average yield for the recruited herds ranged from 4936 to 6644 L of milk per cow per year with a mean herd average yield of 5918 L of milk produced per cow per year.Within our sample population, 144 cows (24%) were in their first lactation, 148 (25%) were in their second, 100 (17%) in their third and 203 (34%) cows were in their fourth lactation and above.There were 235 cows (39%) with MS of 0; 290 (49%) with an MS of 1, denoting imperfect mobility; 61 (10%) with MS 2, denoting impaired mobility; and 9 cows with MS 3 (2%), denoting severely impaired mobility.
Of cows with MS = 0, 43% had 1 or more foot lesions, while 56% of cows with MS = 1; 82% of cows with MS = 2; and all 9 (100%) of cows with MS = 3 had 1 or more foot lesions.
Overall, 54%, 21% and 15% of cows had foot lesions meeting the criteria for case definition 1 (Low), 2 (Moderate) and 3 (High) respectively.Of those cows with 1 or more lesion present, 77% had non-infectious lesions, 9% had an infectious lesion and 14% presented with both an infectious and non-infectious lesion.Of the cows with infectious lesions, digital dermatitis was the dominant infectious pathology, accounting for 96% lesions.Claw overgrowth and white line disease were the 2 most prevalent non-infectious lesions.White line disease was the most prevalent non-infectious lesion of note, accounting for 38% of cases.
Across all case definitions, there was an increase in prevalence of foot lesions by parity, increasing from lactation 1 to lactation 4+ from 0.32 to 0.71, 0.10 to 0.33 and 0.08 to 0.22 for case definitions 1, 2 and 3 respectively.At farm level, prevalence of lesions varied widely, from 0.35 to 0.80 for case definition 1; 0.07 to 0.41 for case definition 2; and 0.04 to 0.31 for case definition 3 (Table 3).
Overall, 12% of cows were positive on MS (MS ≥2).The proportion positive on MS increased with parity, ranging from 0.03 for primiparous cows to 0.23 in cows in their fourth parity and above.The proportion of MSpositive cows varied between farms and ranged from a low of 0.04 on farm 2 to a high of 0.20 on farm 3 (Table 4).

Sensitivity and specificity
The overall sensitivity and specificity of MS at detecting cows with lesions meeting the criteria for Case Definition 1 was 0.18 (95% CI: 0.14 to 0.23) and 0.96 (95% CI: 0.92 to 0.98) respectively.The corresponding overall positive and negative predictive values were 0.84 (95% CI: 0.74 to 0.91) and 0.50 (85% CI 0.49 to 0.51) respectively.The overall sensitivity and specificity of MS at detecting cows with lesions meeting the criteria for Case Definition 2 was 0.35 (95% CI: 0.27 to 0.44) and 0.94 (0.92 to 0.96) respectively, with corresponding positive and negative predictive values of 0.61 (95% CI: 0.50 to 0.71) and 0.84 (95% CI: 0.83 to 0.86).The overall sensitivity and specificity of MS at detecting cows with lesions meeting Case Definition 3 was 0.43 (0.32 to 0.54) and 0.94 (0.91 to 0.95), with corresponding positive and negative predictive values of 0.51 (95% CI: 0.40 to 0.62) and 0.89 (95% CI: 0.88 to 0.90) respectively.
The sensitivity of MS for detecting any lesions (Case definition 1) varied substantially by parity, ranging from 0.04 in primiparous cows to 0.28 in cows of parity 4 and above.Sensitivity of MS remained low in primiparous cows (less than 0.10) across all 3 case definitions.Specificity remained at 0.98 across the 3 case definitions in primiparous cows.There was a substantial increase in sensitivity of MS for the detection of lesions meeting the criteria for case definition 1 in the fourth parity and above cows, with a sensitivity of 0.28.This increased to 0.56 for the detection of lesions meeting the criteria for case definition 3. Conversely, specificity was lowest in this parity cohort decreasing from 0.88 for the detection of lesions meeting case definition 1, down to 0.86 for lesions meeting the criteria for case definition 3.In the case of multiple lesions being identified in the same cow, the cow was assigned the category associated with the highest scoring lesion. 2 Excluding minor lesions expected to have a low probability of impacting gait 3 Only including lesions most likely to result in a detectable gait abnormality

DISCUSSION
Our study has demonstrated that MS threshold of ≥2 has a very low sensitivity coupled with a high specificity for the presence of visible foot lesions and that sensitivity increases with increasing severity of case definition without any detriment to specificity.
We found that 12% of cows had a MS ≥2 varying from 4% to 20% between herds.This finding is similar to previous Irish work: Somers et al. (2015) found 11-15% of cows identified as lame at different times relative to breeding, while Browne et al. (2022b) found a cow-level lameness prevalence of 9%.In contrast, O'Connor et al. (2020a) reported 38.2% of their sampled population of cows as having imperfect mobility, however this was based on a MS ≥1.The prevalence found in our  study is also similar to international work from similar production systems.In New Zealand, Fabian et al. (2014) reported a mean herd lameness prevalence of 9% (based on MS ≥2), with a within-herd prevalence ranging from 1.2% to 36%.In contrast, estimates from indoor production systems appear to be higher (Olmos et al., 2009;O'Connor et al., 2020a;Solano et al., 2015).Similar to previous studies across productions systems, we found that the proportion of cows positive on MS increased with parity (O'Connor et al., 2020a;Browne et al., 2022b;Solano et al., 2015;Jewell et al., 2019).Several explanations for the increase in lameness prevalence in older parity cows have been proposed.Given that all cows in the herd are exposed to a similar set of management and environmental exposures, the number of lameness 'events' is likely to increase as animals age given the cumulative impact of these exposures.These events may be associated with incomplete recovery and chronic changes to the foot (Newsome et al., 2017;Randall et al., 2018).
On foot examination, a lower proportion of infectious foot lesions were found relative to non-infectious lesions in this study.Similar to other studies from pasturebased production systems, white line disease was the most significant non-infectious lesion accounting for 38% of cases (Chesterton et al., 2008;Somers and O'Grady, 2015;O'Connor et al., 2019).
Our study used foot lesions as an outcome against which the Se and Sp of MS was assessed.Importantly, we found that while only 12% of cows had MS ≥2, the proportion of cows with foot lesions on visual inspection was 54% (Case definition 1), 21% (Case definition 2) and 15% (Case definition 3).Accordingly, we found that the Se of MS was surprisingly low, ranging from 0.18 to 0.43 depending on the case definition used.
While few studies have taken the approach we adopted; our findings are consistent with previous observations.For example, Dyer et al. (2007) found that over 37% of foot lesions, measured to be painful, were found on cows with a normal (Sprecher-based) locomotion score, while Manske et al. (2002) found that although 72% of dairy cows examined had at least one foot lesion, only 5.1% were found to be lame based on a locomotion score.Of particular note from our study is that even when restricting the case definition to cows with the most severe lesions that could be expected to have the largest impact on gait (Case definition 3), only 43% of cows were detected.
Of note from our findings is the identification of a significant population of animals that are lesion positive, yet not detected using a MS threshold of ≥2 .While previous studies have shown that at a group level, the average pain index in these animals suggests lower pain than those that are detected on gait obser-vation, substantial levels of pain were still found within this cohort (Dyer et al., 2007).Although there is insufficient evidence at the moment to argue that treatment of these animals would result in improved cure rates, or reduced lesion progression, we argue that the existence and relatively large proportion of animals in this group are a significant cause for concern and a potential area of further research their significance and the role they play in lameness management on farm.
Given these potential welfare risks in these animals, better performing diagnostic tests are required.It is notable that many efforts to develop novel, automated systems have been trained on MS scores rather than foot lesions, which is likely to result in even poorer performance of these automated systems, if detection of treatable lameness-causing foot lesions is the target condition.
The evolutionary pressures to mask signs of pain originating in the wild cattle from which our domesticated species are descended (Huxley and Whay, 20006) may provide and explanation for the poor Se of MS to detect painful foot lesions.One alternative hypothesis could be that visual assessments may rely at least to some degree, in asymmetry of gait associated with unilateral lesions, such that bilateral lesions may not be detected.However, post hoc analysis of our data suggests that this is unlikely to be the case since there appeared to be no difference in the Se of MS when disaggregated according to whether lesions were uni-or bilateral.
Our study used 3 different consensus-based classifications of foot lesions as case definitions against which the Se and Sp of MS was evaluated.In doing so we assumed that different lesion types and severities have different likelihoods of affecting gait, and that this effect is related to the pain experienced by the animal.In our case, we based the case definitions on a consensus view from the co-authors.However, the inclusion of physiological parameters such as, for example nociceptive thresholds (Laven et al., 2008), heart rate and cortisol response (Kotschwar et al., 2009) may have facilitated a more accurate classification of cases.
Our study used multiple definitions of foot lesions as an outcome, against which MS was assessed, i.e., assuming the identification of foot lesions is of interest.This approach should not be confused with an assumption that lesion identification represents a 'better' diagnostic outcome or that the presence of lesions should be considered a gold standard for lameness or pain experienced by the animal.However, as no gold standard test for lameness exists, non-gold standard methods (e.g., latent class approaches), using multiple imperfect diagnostic tests could be useful approaches to consider for future studies.

Logan et al.: LAMENESS IN PASTURE BASED DAIRY COWS
As expected, Se of MS could be improved by reducing the threshold to MS ≥1, increasing to 0.69 (for Case Definition 1).This finding is significant, since previous work has suggested a benefit to treatment of cows with locomotion scores of 2 and greater (similar to MS ≥1 as used in this study) (Schulz et al., 2016).However, this change of threshold was associated with a dramatic decrease in Sp, reducing to 0.49.
In addition, we found that Se also increased according to parity.It is notable that prevalence of lesions also increased by parity.While no formal theoretical relationship exists between sensitivity and prevalence, Brenner and Gefeller (1997) have demonstrated a strong impact of true prevalence on sensitivity owing to the differences in the distributions of underlying traits between populations or cohorts that affect both the prevalence and Se of the diagnostic test.In contrast, the Sp of MS was relatively high, at 0.96, 0.94 and 0.94 for Case definitions 1, 2 and 3 respectively.
A significant strength of this study is that all recruited cows were inspected, irrespective of their mobility score, allowing us to determine the true prevalence of foot lesions.However, while cows were randomly sampled within each herd, herds were recruited from a convenient sample of herds local to our research institution.Our results must therefore be considered with care since it is likely that the distribution of foot lesions in the herd impacts on the Se/Sp of MS as a detection method, and the distribution of these lesions in our herds, may not be reflective of the herds in the country.

Conclusion
Mobility scoring has a poor sensitivity for the detection of cows with foot lesions.Sensitivity could be improved by restricting the case definition to cows with more severe lesion types, however even using this threshold, less than half of the affected animals could be detected with the potential for negative welfare, health and performance implications for affected animals.More effective lameness detection methods with improved Se and Sp are needed for the timely detection of animals with foot lesions, especially if a proportion of these animals are potentially painful and/or would benefit from treatment.
Logan et al.: LAMENESS IN PASTURE BASED DAIRY COWS Table 2. Lesion types and severity and associated case definitions (Y = yes or "present," n = No or "absent" Logan et al.: LAMENESS IN PASTURE BASED DAIRY COWS

Table 1 .
Logan et al.: LAMENESS IN PASTURE BASED DAIRY COWS AHDB Mobility Score Unable to walk as fast as a brisk human pace (cannot keep up with the healthy herd).Lame leg easy to identify -limping; may barely stand on lame leg/s; back arched when standing and walking.Very lame.

Table 3 .
Prevalence (95% CI) of lesions according to case definition by parity and farm 1 Excluding minor lesions expected to have a low probability of impacting gait. 2 Only including lesions most likely to result in a detectable gait abnormality.

Table 5 .
Sensitivity and specificity (95% CI) of mobility score ≥ 2 for detecting foot lesions meeting the criteria for case definitions 1-3 1 Excluding minor lesions expected to have a low probability of impacting gait.2Onlyincluding lesions most likely to result in a detectable gait abnormality