I recently read some posts on Fred Wilson’s blog and it was impressive that he writes every day.
I’ve fallen into the trap of collecting raw material and then waiting to find time to write a 2000-word essay on some topic of importance to me. But, I think it was Steve Jobs who said the best time to do anything was 20 years ago and failing that – best time is now. So now – I will start writing every day and attempt to write on topics of interest to my customers, not me.
We are working on automating patient compliance in medical device clinical trials. Patient compliance is critical for the success of medical device studies.
When we mean success – we mean proving or disproving the scientific hypothesis of the study. Efficacy – is the device an effective treatment for the indication?
Safety – is the device safe for patients?
When we say patient compliance automation we mean the combination of 4 things which depend on each other:
1.Reinforcing patient compliance to the protocol – for example reporting on time and taking the treatment on time. AI-based reinforcement uses data from the patient’s behavior and similar behavior to keep the patient on track without driving him crazy with text or push messaging.
2.Automated monitoring of compliance – using clinical measures and the treatment schedule for the study. An example of a clinical measure is the number of capsules a patient took. An example of treatment schedule is taking the capsules every day before 12.
The output of automated monitoring is real-time alerts and compliance trends to the study team.
3. Automate patient compliance reinforcement using and adaptive control process that takes fresh data from the alerts to make decisions on how to reinforce the patient and keep them on track.
4.In order to automate monitoring and do AI-based reinforcement of patient compliance, you need fresh and up-to-date data.
There is a lot of work being done by startups like Medable, Litmus Health and Flaskdata.io (disclaimer – I am the founder of Flaskdata.io) but it’s a drop in the ocean of 24,000 new clinical trials every year.
Fundamentally – the problem is that the clinical trials industry uses generic solutions developed 40 years ago to assure quality of data-entry from paper forms.
The generic solution used today involves waiting 1-3 days for site data collection to the EDC, and 4-6 weeks for a site visit and SDV and then another 1-12 weeks for a central monitoring operation in your CRO to decide that there was a protocol violation.
You don’t have to be a PhD data scientist to understand that you cannot assure patient compliance to the clinical protocol with 12-week-old data.
The only explanation for using 40-year-old generic solutions is that the CRO business model is based on maximizing billable hours instead of maximizing patient compliance.
It seems that if you want to achieve real-time detection and response and AI-based patient compliance reinforcement, you have to disrupt the CRO business model first.