The formula
One equation. The only system that uses the patient as her own control.
// CORE C2D Drift Scoring Function
// Personal baseline deviation — not population comparison
D_score = |μ_recent − μ_baseline| / σ_baseline
// where:
μ_recent = mean of recent window // e.g. last 90 days
μ_baseline = mean of established baseline // patient's own history
σ_baseline = std dev of baseline values // patient's own variance
// Detection threshold: TH_D = 2.0 standard deviations
// Alternative methods: KL divergence, EWMA, CUSUM change-point detection
if D_score > TH_D: drift_event.seal(feature, time_window, triggering_entries)
// Personal baseline deviation — not population comparison
D_score = |μ_recent − μ_baseline| / σ_baseline
// where:
μ_recent = mean of recent window // e.g. last 90 days
μ_baseline = mean of established baseline // patient's own history
σ_baseline = std dev of baseline values // patient's own variance
// Detection threshold: TH_D = 2.0 standard deviations
// Alternative methods: KL divergence, EWMA, CUSUM change-point detection
if D_score > TH_D: drift_event.seal(feature, time_window, triggering_entries)
Margaret's B12 — 9 years of sealed data
B12 pg/mL · Annual sealed lab panels · 2015–2025
Reference low: 300
Personal baseline: 487
'15
'16
'17
'18
'19
'20
'21
'22
'23
'25
487
Personal baseline 2015
241
Current value 2025
1.94
D-Score · TH_D = 2.0
No population chart flagged this. The value stayed above the reference floor of 300 for eight years. Against her own baseline, the signal was visible in year 4. Metformin was introduced in year 4.
Investor objection
"My doctor already tracks my labs over time. There's a chart in my patient portal."
Why that misses the point
Your doctor's chart compares you to a population. The reference range for B12 is 200–900 pg/mL — a range derived from thousands of patients who are not you. If your personal healthy baseline is 487 and it falls to 241, you are still "in range" by population standards. No alert fires. Your doctor sees a normal value. CORE C2D compares you to your own established baseline using standard deviation units. The same decline that looks normal on a population chart registers as D-Score 1.94 against your personal history. That is the signal. Population charts are blind to it by design.
Same patient. Same data. Two completely different conclusions.
Standard EHR chart · Population comparison
241
✓ Within normal limits
B12 of 241 pg/mL is above the population reference floor of 200. No alert. No flag. Physician sees a normal value and moves on. The 9-year trajectory is invisible in a single column view.
CORE C2D · Personal baseline drift
1.94
⚠ Approaching drift threshold
D-Score of 1.94 against a personal baseline of 487 pg/mL. Decline correlates with metformin initiation in year 4. Current supplementation inadequate. Drift event sealed with triggering entries. Physician briefed before the visit.
0
Prior systems that use the patient's own longitudinal history as the comparison baseline for drift detection — confirmed by full prior art search
"Isn't this just anomaly detection? That's a solved problem."
Anomaly detection compares to a population distribution or a model prediction. Drift detection in CORE C2D compares to the individual's own established personal baseline — the specific D-score formula, sealed as a persistent drift event object with triggering pointers, is what the prior art search confirmed as novel. Our patent attorney rated this cluster HIGH novelty.
"What if the patient has no history? What's the baseline?"
The vault accumulates a baseline as entries seal over time. The system requires a minimum history window before drift scoring activates — a design parameter specified in the patent. This is not a limitation; it is a feature. Every sealed entry makes the baseline more accurate.
The drift engine surfaces what
nine years of normal lab reports buried.
nine years of normal lab reports buried.
Claim Cluster 4 shows how IGFP selects the next question — filling the gaps the drift engine identifies.
Read Claim Cluster 4 →