UforU
Role: Founder & CEO, Website: uforu.co,
Trial and error hurts. UforU uses machine learning to generate a tailored report of the best antidepressants for patients based on their health history. UforU partners with healthcare providers, and also offers its services directly to patients.
Read the research here: Effectiveness of Common Antidepressants.
Article: This Website Could Help You Find the Right Antidepressant on Your First Try.
UforU was created with one purpose in mind: to improve the treatment of depression so that people feel better, faster. It's based on models developed by George Mason University and OptumLabs that are expected to increase remission rates for depression from 33% to 50%:
In addition to improving the lives of people with depression, UforU can help healthcare professionals by offering guidance on which antidepressant a patient should receive, and healthcare organizations by reducing the costs of care and improving health measures such as antidepressant medication management, depression screening and follow-up, utilization of PHQ-9, and depression remission or response.
To get the service into people's hands, I developed a NodeJS website that is deployed to AWS via a Lambda function using their API gateways. AWS offers affordable web hosting that can be easily scaled to meet demand, so it's ideal for a startup with limited cash resources with the potential to grow rapidly. On the site, UforU walks patients through a series of questions to better understand their condition and health history:
The questionnaire zeroes in on the important health data, or predictors, that are used to identify which antidepressants will be most effective for a given person. These predictors allow UforU to predict, for each antidepressant, the likelihood of remission. Here's some sample code that isolates the predictors and searches the model for their presence:
With this information, UforU generates a report, tailored specifically to the patient:
After 3 weeks, we follow up with the patient to see if the recommendations were followed, and again at 14 weeks to see if the recommendations helped, and to what extent.