Скачать или смотреть January 30 - HealthTech Dose

Key Takeaways:
Acknowledge the Lab-to-Clinic Performance Gap: AI diagnostic tools can show high sensitivity (91%) in controlled lab environments but often see significant drops (down to 62%) when faced with the inconsistencies of real-world clinical notes.

Scrutinize Data Quality over Quantity: Large datasets do not guarantee accuracy. In a meta-analysis of over 430,000 Apple Watch users, more than half of the included studies were found to have a high risk of bias.

Look Beyond the "Mean Bias" in Wearables: While a device may have a low average error, the "Limits of Agreement" can reveal fluctuations (e.g., +/- 7 bpm) that are critical for patients with underlying heart conditions.

Modernizing Regulatory Evidence: The FDA is encouraging the use of Bayesian methods, which allow researchers to use prior knowledge to make clinical trials smaller, faster, and more efficient.

Utilize a Systematic "BS Detector": Always evaluate health tech by checking the "Who" (sample diversity), the "Where" (lab vs. real world), the "Limits" (admitted study weaknesses), and the "Data" (transparency and risk of bias).