People argue the most about things they are uncertain about.
People argue the most about things they are uncertain about. They don’t argue about the sun setting or rising.
Zen and the Art of Motorcycle maintenance
I had 2 observations today about digital health and clinical research.
My first observation is that there are too many buzz-words in clinical research. EDC, ePRO, eCOA,IRT, IxRS, eSource, eConsent, CTMS, CDM, SDTM, eClinical and more. This is not a good sign.
Look, I may be an entrepreneur, but I’m also a programmer. In the Node.JS world you have NPM with over 1 million packages. Free, infinite variety for the price of 1 buzzword.
API-based interoperability between the eClinical vendors simply does not exist. API’s are often inaccessible, non-existent or proprietary. A search for Medidata on Github reveals repositories that were last updated over 2 years ago. Most are empty. The most recent Medidata-related repo on Github is a third-party Go language wrapper to the proprietary Medidata authorization protocol. From this perspective, Medidata is doing a great job. If you want to connect to an Oracle Life Science product “To receive a physical media pack with all the required DVDs, contact Oracle Support”.
My second observation this morning was that vanity ensures failure. This is a corollary of not following 3 basic design principles for interactive software systems (there may be more — these are the 3 that have the most empirical data in projects I’ve done over the years).
I suppose programmer vanity and violation of design principles, is the main reason that clinical operations front-liners (CRCs, CRAs) hate the eClinical platforms they (are forced) to use.
Pride goes before destruction, a vain spirit before a fall. Proverbs 16, v18.
The 3 design patterns for interactive software systems
In interactions between people and technology, there are 3 fundamental design patterns that guide us: openness, speed, simplicity.
- Open, vendor-neutral standards enable communication between smart people trying to solve difficult problems.
This is the secret of the success of the Internet. If you have a question, you can Google and find the answer in Stackoverflow; you don’t have to contact Oracle support and ask for a DVD. This is how software developers work today. For clinical and regulatory consultants, free, online accessibility to FDA and EMA guidance and standards is key to their service delivery model. For systems integrators, you want to install an SDK on your MacBook, try it out and then deploy to AWS. You do not want to be accepted into a vendor partner program (like Medidata) (or buy a license and ask for DVDs) (like Oracle) in order to integrate their EDC with your new beautiful IxRS that you developed in React Native.
You want to do the work, put in the beach and see if it survives.
2. Real-time is better than batch-time
The easiest way to understand why fast measurements are better than slow batch processing is to ask yourself a question. Do you remember what you were doing 3 days ago at this time? No. Do you remember what you were doing 3’ ago?. Of course you do.
In a clinical trial, it is easier to fix a problem while the patient is still in the clinic than to wait for a data manager to respond to a query raised by the EDC by the site coordinator data entry 3 days after the patient went home.
Real-time processing is also a question of physics, not just context. You can sip from a glass, not from a fire-hose. A remote study monitor can execute the playbook for an alert for a single patient. A CRA processing 70 patients once/month will be overwhelmed.
3. Simplicity and reduction of moving parts in a clinical trial is important at 2 levels of concern:
The first level of concern is the clinical data model. The most important part of getting a clinical data model right is to resist that ugly temptation to collect lots of additional data unrelated to your hypothesis. Whenever tempted, consider that the most important and elegant scientific experiments in history were the simplest.
Eratosthenes estimated the Earth’s circumference with a stick and 3 assumptions. He measured the shadow cast by a stick in Alexandria at noon of the summer solstice. He found the angle of sunlight there to be 7.2 degrees, or 1/50th of a circle’s 360 degrees.
- The distance between Aswan and Alexandria is 5,000 stadia.
- Aswan is precisely on the Tropic of Cancer; at noon of summer solstice the sun is directly overhead.
- Aswan and Alexandria are on the same meridian
- Knowing that Earth was spherical, Eratosthenes estimated the Earth’s circumference by multiplying the distance between the 2 cities by 50. He arrived at 250,000 stadia, close to the correct figure of 24,900 miles.
The second level of simplicity is in the user experience. For a patient, the best UX is between 0 and 1 interaction. 0 interaction — collect data passively from a phone or wearable device. 1 interaction — click once on a mobile device. For a study nurse, enrolling a patient in 1 click is best.
With a simple data model and minimalistic UX, the margin for error is reduced and our ability to debug the system is improved — since there are less moving parts to break.
As my Dad used to tell us when we were kids, “if you play with it long enough it will eventually break”.