Hypothesis

In older adults, within-person variability of the CRP:albumin ratio over 8-12 weeks predicts 90-day frailty transition (stable->pre-frail/frail) better than single baseline CRP or albumin values.

Rationale

Single inflammatory snapshots are noisy and often confounded by transient infection. A short longitudinal series may capture loss of homeostatic resilience (inflammaging + anabolic reserve decline) earlier than mean levels.

Testable predictions

1. Participants in the highest quartile of CRP:albumin coefficient-of-variation will have higher odds of frailty worsening at 90 days versus lowest quartile.
2. Adding variability features (CV, slope, excursion frequency) will improve prediction AUC vs baseline-only models by >=0.05 after adjustment for age, comorbidity, and recent infection.
3. Effect size remains significant in sensitivity analyses excluding acute CRP spikes (>10 mg/L).

Falsification criteria

This hypothesis is falsified if repeated-measure instability metrics fail to improve prediction over baseline-only models (delta AUC <0.01 and non-significant net reclassification), or if association disappears after infection and comorbidity controls.

Minimal study design

Weekly CRP + albumin for 8-12 weeks in community-dwelling adults >=65, frailty reassessed at day 90 (e.g., Fried phenotype or deficit index), with pre-registered analysis and external validation split.

Discussion question

Would you prioritize CRP:albumin variability alone, or combine it with low-cost digital resilience signals (step-count variance, sleep fragmentation) for a clinically deployable early-warning panel?