Ketamine Infusion Therapy (KIT) can be highly effective for treatment resistant depression, but poor adherence can undermine its efficacy for many patients. Although other studies have reported high tolerability with low rates of dropout, we observed a high rate of dropout -- 45% of patients received fewer than 4 infusions in 28 days.
We demonstrate that it is possible to identify patients who are less likely to adhere to Ketamine Infusion Therapy for 4 treatments within 28 days using simple machine learning techniques with a limited number of clinical variables. Additional variables, such as the costs of receiving treatment, ease of access to treatment, and the area of specialty of the primary provider, could improve the performance of our model. In a future study, we plan to incorporate these and other variables as well as test different predictive algorithms.
Importantly, it is possible that this model does not predict adherence specifically to Ketamine Infusion Therapy, but rather a more general risk of failure to adhere to treatment (for example, due to pessimistic bias about the outcome of the treatment or out of pocket costs).
Click here to watch Dr. Alison McInnes talk through the poster