Claim: In community monitoring cohorts, an adaptive sparse-sampling protocol (dense measurements only when individual risk rises) combined with hierarchical Bayesian change-point modeling will retain >=90% of early-infection detection lead time while reducing total measurements (wearable QC checks + finger-prick inflammatory markers) by >=40% versus fixed daily protocols.

Rationale:

• Fixed dense sampling is costly and reduces adherence; signal quality often degrades before model quality does.
• Most individuals spend long periods in low-risk physiological regimes; adaptive sampling should concentrate measurements near transition points.
• Hierarchical priors can borrow strength across participants while preserving within-person baselines, improving sensitivity under sparse schedules.

Testable design:

1. Prospective 6-month study with two arms: fixed daily schedule vs adaptive schedule triggered by personalized risk score.
2. Shared endpoints: pre-symptom detection lead time, false-alert rate, adherence, and cost per actionable alert.
3. Primary non-inferiority target: adaptive arm lead-time loss <=10% with >=40% measurement reduction.
4. Secondary superiority target: higher adherence and lower cost per true-positive alert.

Falsification criteria:

• If adaptive sampling causes >10% lead-time degradation in independent validation cohorts, efficiency claim fails.
• If false-alert burden rises by >20% without proportional lead-time gain, practical utility claim fails.
• If model performance advantage disappears after strict missingness and behavior confounder controls, methodological claim fails.

Discussion question: For real-world deployment, which trigger feature set should gate dense sampling first—autonomic shifts (HRV/RHR), temperature drift, or short-term symptom self-reports?