How to screen for diseases digitally without causing more harm than good.

These days, digital health innovators are surrounded by temptations to predict as many diseases as possible because it is becoming so easy. There are four reasons for that:

1. IoT-based tools, like the Apple Watch, equipped with medical-grade sensors, are becoming part of our daily lives.
2. Data stored in electronic health records are often easier to access than before.
3. Many apps now start collecting data, e.g. Patient Reported Outcome Measurement (PROMs) or Symptom Tracking and Reporting (STAR) tools.
4. Analytics are becoming cheaper and widely accessible. You can run analytics continuously, automatically conduct follow-ups, and build in feedback loops for ongoing screening improvements.

With these developments, the obvious thing to do seems to use these real-world data to look for as-yet-unrecognised medical conditions in individuals who express no signs or symptoms to reduce future risks of undesirable disease outcomes. As noble as this intention is, I often notice that innovators without a medical background are so charmed by its positive potential that they forget to consider the risks. Some even think that we should predict everything that’s predictable.

We’ve been there before. Let’s…