Definition
OEI measures the ratio of minimum to maximum outreach intensity across demographic subgroups targeted by AI-driven patient engagement systems. An OEI of 1.0 means all subgroups receive identical outreach intensity; values approaching zero indicate that some populations receive dramatically less engagement effort than others. The metric exposes whether algorithms optimizing for response rates or cost efficiency systematically deprioritize populations that are harder to reach, reinforcing existing health disparities through differential attention.
The Measurement Gap
Measurement Gap and Decision
Decision Question: Does the AI-driven outreach system engage all patient populations with equitable intensity, or does it concentrate resources on populations most likely to respond. Amplifying existing access disparities?
What Goes Unmeasured Today:
Patient outreach systems report aggregate metrics: total contacts made, response rates, appointment conversions. These aggregate numbers mask profound distribution inequities. An AI system optimizing for response rate will learn, correctly, that certain populations respond at higher rates, and will allocate more outreach resources to those populations. Patients with stable housing, reliable phone numbers, digital literacy, and English fluency are easier to reach and more likely to respond. The algorithm rewards efficiency by concentrating effort where returns are highest. Meanwhile, populations with lower response rates (often those with the greatest health needs) receive progressively less outreach. The system reports improving overall engagement metrics while systematically abandoning the patients who need engagement most. No current metric captures this distribution. A health system can report record outreach volumes while entire demographic subgroups receive a fraction of the contact intensity directed at easier-to-reach populations.
Invisible Failure Mode:
Algorithmic optimization of outreach creates a feedback loop that widens health disparities. High-responding populations receive more outreach, respond more, and receive even more outreach. Low-responding populations receive less, disengage further, and disappear from the system's attention. The aggregate metrics improve while equity deteriorates, and without measuring outreach distribution by subgroup, the disparity is invisible.
Formula
Primary Formula:
OEI = min(Outreach_Intensity_subgroup) / max(Outreach_Intensity_subgroup)
Where Outreach_Intensity = total outreach contacts per eligible patient within each demographic subgroup over the measurement period.
Variable Definitions:
| Variable | Definition | Source |
|---|---|---|
| Outreach_Intensity_subgroup | Total outreach contacts (calls, texts, mailings, portal messages) divided by eligible patients within each defined subgroup | Outreach platform + patient registry |
| Demographic subgroups | Population segments defined by race/ethnicity, language, age band, insurance type, ZIP code, or other equity-relevant dimensions | Patient demographics (EHR/ADT) |
| min / max | Minimum and maximum outreach intensity values across all defined subgroups | Calculated |
Data Requirements:
Outreach contact logs with patient identifiers and contact type from outreach platform. Patient demographic data including subgroup classification dimensions. Eligible patient denominators by subgroup from patient registry or panel management system.
Edge Cases and Limitations:
Subgroup definitions determine what the metric sees. OEI calculated by race may miss disparities by language or geography. Report across multiple segmentation dimensions. Very small subgroups produce unstable intensity calculations; set minimum subgroup size thresholds. Outreach intensity is not outreach effectiveness: equal contacts do not guarantee equal engagement.
Proposed formula, pending validation pilots.