Predicting Our Way Out Of A Crisis
The belief that we can “Minority Report” our way out of a crisis of overdose deaths is not only wrong, it’s deeply dystopian.
A new product called AvertD has recently been approved by the FDA.
As an emergency and addiction medicine physician, this worries me a lot, actually.
These developments can have disastrous unintended downstream consequences and can very easily lead to increasing stigma and under-treating pain. We cannot adequately predict future outcomes, even with genetic testing, and I don’t believe tools like this have significant validity. As noted on MedPage Today, the company which created the test, SOLVD, presented results of a single observational study that utilized machine learning as a predictive algorithm to the FDA which demonstrated a sensitivity of 82.76% and a specificity of 79.23% for detecting opioid use disorder. This level of specificity for what should be considered a critical clinical decision instrument would be considered unacceptable by most standards for predicting bad outcomes.
For example, we use risk stratification clinical decision instruments all the time in emergency medicine to aid our decision making. Do you think discharging someone having chest pain with less than 80% certainty that…
