Back to Reference Library
farriery
veterinary
biomechanics
anatomy
nutrition
physiotherapy
2023
Systematic Review

Clinical predictive models in equine medicine: A systematic review.

Authors: Cummings Charles O, Krucik David D R, Price Emma

Journal: Equine veterinary journal

Summary

# Editorial Summary Predictive clinical models—mathematical tools that forecast outcomes such as survival or surgery requirement based on a horse's baseline data—remain poorly characterised within equine medicine, despite their potential to refine decision-making. Cummings and colleagues conducted a comprehensive systematic review using PubMed and Google Scholar to identify all published multivariable predictive models for equine patients, assessing their methodological quality using the PROBAST (Predictive model Risk of Bias Assessment Tool) framework and evaluating their external validation and clinical applicability. Of 90 models identified across the literature, colic-related predictions dominated the field at 41%, yet all included models exhibited high risk of bias—primarily stemming from analytical weaknesses rather than problems with their underlying study populations—though this does not render them clinically unusable without further scrutiny. Concerns regarding applicability were reassuringly low for most models, suggesting that where bias is addressed, many could feasibly transfer to different clinical settings and patient populations. For equine professionals involved in clinical decision-making, this review underscores both the growing availability of predictive tools and the critical need to evaluate their methodological rigor before implementation; whilst colic prediction dominates current research, expansion into other conditions and improved validation practices across the discipline would substantially strengthen their utility in practice.

Read the full abstract on PubMed

Practical Takeaways

  • Clinical predictive models for horses exist and may help decision-making, particularly for colic cases, but require careful scrutiny before clinical adoption
  • High bias risk in published models does not automatically exclude them from use, but warrants cautious interpretation and awareness of their limitations in your specific patient population
  • Before implementing any predictive model in practice, verify how the model was validated and whether its patient population matches your case types

Key Findings

  • 90 predictive models and 9 external validation studies were identified across equine medicine literature
  • 41% of models were developed to predict colic-related outcomes such as need for surgery or survival
  • All included models were classified at high risk of bias, primarily due to analysis-related methodological issues
  • Applicability concerns were low for the majority of models despite high bias risk

Conditions Studied

colicsurgical colicoutcomes requiring surgerysurvival to discharge