01

Current progress and boundaries

Health apps and wearables can flag trends, but data quality varies across devices and long-term adherence often declines.

02

Why it remains difficult

Sleep, diet, medication, and social support affect one another. A model must distinguish measurement error and temporary fluctuations from genuine changes in risk, while accounting for whether patients can follow the advice.

03

Paths worth exploring

Begin with one condition and a clearly defined care pathway. Limit recommendations to reviewed protocols, and design patient confirmation, care-team review, and escalation for unusual cases.

04

How to evaluate success

Track follow-up completion, alert accuracy, the share escalated to people, and patient burden. Assess health outcomes over an appropriate period rather than relying on click-through rates.