Some service dogs are trained to give an alert associated with a medical condition. The task, evidence, and reliability need to be evaluated for the individual partnership.
Keep the evidence specific
Research on dogs trained to alert to out-of-range blood glucose found variable performance. An alert should prompt the person to follow their established monitoring and care plan, rather than rely on the dog as a substitute for medical measurement.
Seizure-response tasks and advance seizure prediction are different. The Epilepsy Foundation cautions that advance-prediction claims need careful scrutiny and more research. A trainer cannot guarantee that any dog will learn to predict seizures.
Broad claims that dogs can reliably smell all allergens, depression, or future medical episodes require evidence beyond anecdotes. The earlier article made those claims without adequate support; they have been removed.
Build the alert behavior safely
Work with appropriate medical and training professionals on whether the proposed task is suitable and how it will be evaluated. Use reward-based methods to teach a clear response. Plan for missed alerts and false alerts, and evaluate performance beyond a single controlled exercise.
Do not trigger a medical episode to create a training scenario. Collection or storage of biological samples needs an appropriate safety protocol. Voluntary registration does not establish medical accuracy or legal service-dog status.
Sources
- Rooney and colleagues: Variable performance of glycaemia-alert dogs
- Epilepsy Foundation: Seizure response and prediction
- AVSAB: Reward-based training
- DOJ: Task definition and registry documents