At a glance
A major US healthcare system needed analytics that went beyond “who is at risk” to answer the harder question: who is at risk and amenable to intervention, in ways that close — rather than widen — equity gaps, and where the underlying utilization is actually avoidable? UNVEIL was retained as a direct external contractor to deliver this stream of work alongside the population’s predictive risk-modeling effort. Findings were designed to inform the client’s value-based-care contracts, health-equity initiatives, and supportive-housing strategy.
The situation
The client operates in a value-based-care environment where being penalized for unavoidable utilization is bad analytics, and intervening with patients who would not benefit is bad medicine. Three connected questions had to be answered together to allocate care-management resources well:
- Impactability — among the highest-risk patients, which ones have plausibly preventable events that an outpatient intervention could change?
- Equity — where do outcomes, risk levels, or intervention amenability diverge across subgroups defined by social and demographic factors? Does the program reduce or reproduce those gaps?
- Cost / avoidable utilization — which slice of the predicted utilization is genuinely avoidable, and what is its dollar shape under the system’s risk-bearing contracts?
The strategic context spanned population health, health equity, value-based care, financial sustainability, and supportive-housing initiatives — and the analyses had to support all five.
The challenge
These three lenses interlock and contradict each other in interesting ways:
- Risk ≠ impactability. A frail patient near end-of-life is high-risk, but their hospitalization may not be preventable in any clinically or ethically meaningful sense. Targeting them for “impactability” interventions wastes resources and can harm care.
- Optimizing against avoidable cost can entrench inequity. If the most “impactable” patients in your data are the ones the existing system already serves well, doubling down on them widens gaps. Equity analysis has to be in the loop, not added as an afterthought.
- Social determinants are signal, not noise. Housing instability and homelessness are predictors of avoidable utilization in their own right, and targets for upstream intervention. Both framings had to be supported.
- Healthcare-grade rigor. Calibration across subgroups, leakage controls, sensitivity to small-n cohorts, and clear documentation of methodological choices — non-negotiable for clinical and contracting use.
Our approach
We worked inside the client’s secure cloud ML environment, using their data and infrastructure standards.
- Impactability framing. Translated the published literature on predictive impactability — the move from raw risk to amenability to intervention — into operational definitions grounded in the client’s data: avoidable ED visits, avoidable inpatient stays, long lengths of stay, and 30-day all-cause readmissions as the canonical impactable-event categories.
- Subgroup-aware risk lenses. Built risk views that reported performance, calibration, and lift across demographic and SDOH-relevant subgroups — so leadership could see whether a given strategy reduced or widened equity gaps before it was deployed.
- Social-determinants modeling. Developed risk views for housing instability and likelihood of becoming homeless, treating these both as predictors of avoidable utilization and as outcomes worth predicting in their own right — supporting the system’s supportive-housing strategic initiatives.
- Avoidable-utilization framing. Mapped predicted utilization to the avoidable-event categories defined above, with documentation of the assumptions, the populations they applied to, and the boundaries of clinical/operational interpretation.
- Documentation and handoff. Produced analysis documents covering definitions, methods, results, limitations, and operational guidance — sized for the in-house analytics team and for use in conversations with clinical leadership and value-based-care contracting.
The outcome
- An analytic foundation the client could carry into care-management prioritization, value-based-care contracting, and health-equity strategy conversations.
- Subgroup and SDOH views that made equity questions explicit and answerable inside the same framework as risk and avoidable utilization — rather than in a parallel, easily-deprioritized track.
- A clean handoff: the client owned every artifact (code, documentation, methods notes), with no dependency on UNVEIL for ongoing analysis or extension.
What this means for you
If you operate a value-based-care program, a population-health initiative, or any decision-support system where you need to allocate finite intervention resources well, we can:
- Move the conversation from risk to impactability — focus on who actually benefits from intervention.
- Bake health-equity diagnostics into the same framework as risk and cost — so equity is not a separate track that loses out to the operational dashboards.
- Quantify avoidable utilization in terms your contracting and finance teams can act on under risk-bearing arrangements.
- Treat social determinants as both predictors and outcomes worth modeling — supporting upstream investment in housing, food, and other SDOH levers.
- Deliver everything as documented, transferable analyses your in-house team owns and can extend.
Want to talk about a similar problem in your population? Contact us.