The use of collar-worn accelerometers represents a major innovation for the objective monitoring of pruritic skin conditions in dogs.
This emerging technology offers veterinary practitioners a reliable means to quantify pruritic behaviors and evaluate the effectiveness of prescribed treatments, thus filling a significant gap in the management of canine dermatoses.
Context and Problematic
Pruritic skin conditions are a frequent reason for consultation in canine veterinary medicine. Pruritus, a major clinical manifestation of many dermatoses such as atopic dermatitis, represents a considerable diagnostic and therapeutic challenge for practitioners. Unlike other medical conditions whose diagnosis relies on standardized procedures, the identification and management of the underlying causes of pruritus require a multifactorial approach integrating anamnesis, observation of clinical signs, and monitoring of therapeutic response.
Traditional evaluation of pruritus intensity primarily relies on subjective owner observations, which often limits the accuracy of assessing treatment efficacy. This subjectivity constitutes a major obstacle in optimizing the therapeutic management of canine pruritic dermatoses.
Chronic dermatoses like atopic dermatitis impose a significant burden both on the animal – which suffers from persistent pruritus, erythema, and discomfort – and on the owner, who faces recurrent veterinary consultations, continuous medication expenses, and sometimes the implementation of specific diets. This situation highlights the importance of developing objective and precise methods to evaluate and monitor the evolution of canine pruritus.
Better monitoring of pruritus will optimize current protocols, especially in atopic dermatitis.
Contribution of Objective Monitoring Technologies
The advent of collar-worn accelerometers offers promising prospects for objective monitoring of pruritic behaviors in dogs. These devices, equipped with deep learning algorithms, can accurately identify specific behaviors such as scratching and licking, thus providing objective quantification of pruritus intensity in the animal’s daily environment.
Previous studies have validated the effectiveness of these accelerometers for detecting pruritic behaviors, with sensitivity ranging from 76.9% to 87.0% and specificity reaching 99.8% for scratching, as well as sensitivity of 77.2% and specificity of 99.0% for licking. Significant concordance has also been demonstrated between accelerometer-recorded scratching measurements and those reported by owners via a visual analog scale.
This study is part of this innovation dynamic by exploring the application of collar-worn accelerometers for objective monitoring of the efficacy of dermatological treatments in canine pruritus. The main objective is to determine if these devices can effectively detect and quantify behavioral variations induced by different therapeutic modalities under real-world veterinary practice conditions.
Methodology
Study Design
This retrospective observational research is based on data collected between 2019 and 2023 as part of the “Pet Insight Project,” an initiative that distributed 100,000 collar-worn accelerometers (Whistle FIT; Mars Petcare) to clients of Banfield veterinary hospitals (a network of primary care clinics operating in over 1,000 locations across 42 US states).
Study Population and Data Collection
This analysis focused on 1,803 dogs with a diagnosis of dermatitis in their electronic medical record (EMR). For each animal, a random episode of dermatitis was selected so that each dog was represented only once in the dataset. A subset of 864 dogs that received only one type of therapeutic intervention was also formed to allow for the identification of effects directly attributable to specific interventions.
The analyzed data included:
- EMRs collected during consultations for dermatological conditions at Banfield veterinary hospitals
- Daily measurements of pruritic behaviors (licking and scratching) obtained through the collar-worn accelerometer
Pruritic behaviors were identified by FilterNet, a deep learning algorithm processing accelerometer data, with validated performance for detecting licking (sensitivity: 0.772; specificity: 0.990) and scratching (sensitivity: 0.870; specificity: 0.997).
Inclusion and Exclusion Criteria
Dogs were included if they had a diagnosis of “dermatitis” or “atopic dermatitis” in their EMR and had made at least two visits to a Banfield veterinary hospital. Visits were filtered to include only diagnoses established on the date of consultation, thus excluding historical mentions of dermatitis.
To ensure the robustness of behavioral measurements, dogs had to have at least 20 valid days of accelerometer recordings during the 30 days before and after their dermatological consultation (i.e., 60 days of evaluation in total). A recording was considered valid if the accelerometer had collected data for at least 12 hours out of 24. Missing data were imputed using a personalized model for each dog.
Categorization of Therapeutic Interventions
Therapeutic interventions were extracted from medical notes and invoices present in EMRs, then classified into different categories:
- Antimicrobials
- Antibodies (biological immunotherapy)
- Antihistamines
- Antiparasitics
- Immunosuppressants
- Medicated shampoos
- Analgesics
- Anti-inflammatories
A control group of animals with no recorded intervention was also formed to allow for comparisons.
Statistical Analysis
The analysis was performed in Python (version 3.9.13) using the pandas (version l.3.5) and statsmodels (version 0.14.0) libraries.
To evaluate the overall effect of therapeutic interventions, a linear regression modeled pruritic behaviors in the post-intervention period. Baseline was defined as the average time spent on the studied behavior in the 14 days preceding the intervention. The primary outcome was the average percentage change in pruritic behavior in the 30 days following consultation relative to baseline.
A Directed Acyclic Graph (DAG) was constructed to illustrate the hypothetical scenario and identify potential confounding factors. The number of previous consultations related to dermatitis was identified as a confounding factor and therefore controlled for in the statistical model.
To differentiate the effect of various interventions, a linear regression model was applied to the subset of dogs that received only one type of intervention, with the presence/absence of a specific intervention identified in the EMR as the exposure variable.
Results
Characteristics of the Study Population
The main cohort included 1,803 dogs, of which 55% were male and 45% female, with a high proportion of sterilized animals (96%). The average age was 6.2 years (standard deviation: 3.2). The subset of animals that received only one type of intervention (n=864) had similar demographic characteristics.
The analysis of pruritic behaviors showed that the average daily time spent licking was 1,038.9 seconds (standard deviation: 918.1) before consultation and decreased to 899.6 seconds (standard deviation: 832.9) after intervention. For scratching, these values changed from 105.2 seconds (standard deviation: 123.2) to 82.4 seconds (standard deviation: 101.9).
Temporal Evolution of Pruritic Behaviors
Analysis of the chronology of pruritic behaviors revealed a progressive increase in time spent scratching and licking in the days preceding dermatological consultation, followed by a decrease and stabilization at a level lower than baseline after therapeutic intervention. This temporal evolution confirms the ability of accelerometers to detect behavioral fluctuations associated with pruritic episodes and treatment effects.
Overall Effect of Therapeutic Interventions
The linear regression model demonstrated a significant effect of therapeutic interventions on scratching behavior. The coefficient related to the intervention was -25.6 ± 4.43 (95% CI: -34.3 to -16.9; p < 0.005), indicating a substantial reduction in scratching following interventions. Considering the average number of previous consultations for dermatitis, the predicted average reduction in scratching was 16.6% in dogs that received an intervention, compared to an increase of 9.0% in those with no recorded intervention.
For licking behavior, the intervention coefficient was -8.6 ± 5.62 (95% CI: -19.60 to 2.46; p = 0.13), suggesting a trend towards reduction that did not reach the threshold of statistical significance. The predicted average variation in licking was -7.9% with intervention versus +0.7% without intervention.
Differentiation of Effects by Intervention Type
Analysis of the specific effects of different intervention categories revealed substantial variations in their impact on pruritic behaviors:
-
Antibody treatments (biological immunotherapy): Most significant reduction in scratching with a coefficient of -48.12 ± 5.70 (95% CI: -59.32 to -36.93; p < 0.005), corresponding to an average decrease of 35.6% (95% CI: -42.22 to -27.89) in scratching behavior relative to baseline.
-
Immunosuppressive treatments: Significant effect on scratching with a coefficient of -30.62 ± 7.36 (95% CI: -45.06 to -16.19; p < 0.005), representing an average reduction of 17.6% (95% CI: -29.48 to -5.64).
-
Medicated shampoos: Significant decrease in scratching with a coefficient of -22.94 ± 9.78 (95% CI: -42.13 to -3.74; p = 0.02).
-
Anti-inflammatories and antimicrobials: Less marked effects on scratching, with coefficients of -24.95 ± 15.44 (p = 0.11) and -6.68 ± 5.93 (p = 0.26) respectively.
For licking behavior, immunosuppressive treatments (-20.36 ± 6.95; p < 0.005) and antibody treatments (-17.98 ± 5.39; p < 0.005) showed the most significant reductions, while antiparasitics were associated with a paradoxical increase (15.77 ± 7.68; p = 0.04).
Influence of Chronicity on Treatment Efficacy
A particularly interesting result concerns the interaction between the efficacy of interventions and the number of previous consultations for dermatitis. For scratching behavior, the therapeutic effect decreased with an increasing number of previous consultations, suggesting a lower response to treatments in chronic or recurrent cases.
This relationship was particularly evident for medicated shampoos, whose efficacy decreased from an average scratching reduction of 15.7% (95% CI: -33.00 to 1.56) at the first consultation to only 1.1% (95% CI: -20.09 to 17.89) at the sixth. This association between chronicity and therapeutic response was not observed for licking behavior.
Discussion
Study Contributions to Clinical Practice
This study represents the first large-scale retrospective analysis exploring the application of collar-worn accelerometers for objective monitoring of pruritic behaviors in dogs under real-world conditions. The results clearly demonstrate the ability of these devices to detect and quantify behavioral variations induced by different therapeutic modalities in the management of canine pruritic dermatoses.
The identification of a substantial reduction in scratching behaviors following therapeutic interventions, particularly after the administration of antibodies (40.9% reduction) and immunosuppressants (23.4% reduction), confirms the clinical utility of accelerometers as tools for objective evaluation of dermatological treatment efficacy.
For veterinary practitioners, this technology offers an invaluable opportunity to remotely and objectively monitor the therapeutic response of patients suffering from pruritic skin conditions. By reducing the subjectivity inherent in traditional evaluations based on owner observations, accelerometers allow for a more nuanced understanding of clinical evolution and promote the development of personalized and proactive therapeutic approaches.
Methodological Considerations and Limitations
Several factors must be considered in interpreting our results. The efficacy of therapeutic interventions depends on many parameters, including treatment adherence, which could not be evaluated in this retrospective study. For treatments administered by owners, it is possible that the application was not always optimal, which could have influenced the observed results.
The decreased efficacy of interventions in dogs with a higher number of previous consultations for dermatitis raises several hypotheses. This observation could reflect the inadequacy of prescribed treatments for certain chronic conditions or the presence of unidentified confounding factors, such as secondary infections complicating chronic pruritus cases.
This study also highlighted a notable difference between the effects of therapeutic interventions on scratching and licking behaviors. This disparity could be explained by the specific association between certain types of dermatitis and the location of pruritus, as well as by the versatile nature of licking behavior, which can respond to stimuli other than pruritus (grooming, anxiety, etc.).
Although this analysis compared pruritic behaviors of dogs that received documented intervention to those without recorded intervention, it is possible that some animals benefited from treatments not documented in their EMR. This potential limitation would, however, tend to diminish the observed differences between groups, suggesting that the actual effects could be even more pronounced than those reported in this study.
Future Perspectives
The results of this research open promising perspectives for the integration of accelerometers into veterinary dermatological clinical practice. These devices could be a valuable complement to traditional evaluation methods, offering veterinarians and owners an objective tool to optimize the management of chronic pruritic dermatoses.
Future studies could explore in more detail the factors influencing the efficacy of different therapeutic modalities and elucidate the mechanisms underlying the variability of response based on the chronicity of the condition. Integrating complementary data on therapeutic adherence, non-medical interventions, and the initial severity of pruritus would refine our understanding of care pathways for dogs with chronic pruritic dermatoses.
Technological developments aimed at evaluating not only the duration but also the intensity of pruritic behaviors could also enrich the informative value of accelerometers and further improve the precision of therapeutic monitoring.
Conclusion
This retrospective study demonstrates that collar-worn accelerometers are reliable tools for the objective detection and quantification of behavioral variations induced by different therapeutic modalities in the management of canine pruritic dermatoses.
The analysis of data collected from 1,803 dogs under real conditions allowed for the identification of significant reductions in scratching behaviors following therapeutic interventions, with particularly marked effects for antibody and immunosuppressive treatments. The ability of accelerometers to differentiate the impact of various intervention categories underlines their potential as decision-making tools in the field of veterinary dermatology.
By offering an objective and non-invasive method for remote monitoring of therapeutic efficacy, these devices could contribute to optimizing the management of chronic pruritic dermatoses and significantly improve the quality of life of affected dogs as well as owner satisfaction.
O’Rourke A, Redford L, Carson A, Lyle S, Kresnye C, O’Flynn C. Retrospective observational study shows accelerometers can monitor effects of canine pruritus treatment. Am J Vet Res. 2025 Jan;24(09):1-8. doi:10.2460/ajvr.24.09.0269.
