Define whose health counts
The first strategic choice is the denominator. A health plan’s membership, an ACO’s attributed lives, a provider’s active patients and a community’s residents are different populations. A program can improve the economics of one while leaving the needs of another unaddressed.
I would make those boundaries explicit before selecting an intervention. Who is eligible, who is missing from the data, who can actually receive the service and who remains accountable when coverage or attribution changes? Public-health leadership has to see beyond the people who have already reached a reimbursable encounter.
The public-value question is also broader than medical savings. Better function, more dependable access, reduced caregiver burden and fewer barriers to participation can matter even when a short-term claims analysis does not show a positive financial return. The funding case must state which objective it is buying.
Allocate by need and plausible response
A high risk score predicts an outcome under a model. It does not establish that a specific intervention can change that outcome. I separate three judgments: severity of need, plausible benefit from the proposed service and feasibility of delivery. A fourth judgment concerns the people whose need is least visible in the available data.
This changes the resource decision. An additional navigator, transport contract, primary-care session, behavioral-health appointment or home-based service addresses a different constraint. Funding more coordination when no receiving capacity exists can produce an efficient queue of unmet need. Funding capacity without an accessible path into it can leave the investment unused.
Make the portfolio multi-objective
| Decision layer | Question to answer | Measure to preserve |
|---|---|---|
| Population benefit | Which health, function or access outcome should improve? | A defined population, endpoint and follow-up period. |
| Equitable reach | Who is eligible but never reached or served? | Progression by relevant population group, with data-quality checks. |
| Delivery capacity | Which service can change the outcome, and is it available? | Accepted referrals, completed services and unresolved need. |
| Economic sustainability | Who funds delivery and who receives the benefit? | Fully loaded program cost, purchaser economics and fiscal spillovers. |
| Learning | What result would change resource allocation? | Comparator, uncertainty and a documented decision threshold. |
These objectives cannot always be reduced to a single dollar score. I would establish minimum access and quality standards, identify unacceptable distributional effects and then compare the incremental benefit of alternative investments. A program that improves the average while widening a serious access gap requires redesign, even when the financial case looks attractive.
Contract across institutional boundaries
The organization paying for an intervention may not receive the benefit. A provider may absorb coordination cost while a payer captures lower acute spending; a community organization may carry the delivery burden while another institution receives a performance payment. Unresolved, this becomes an underinvestment problem.
I would map the beneficiary of each material effect, the timing of that effect and the institution able to fund the enabling work. Prospective support, service contracts and shared investment may be needed before an outcome-based component becomes credible. Community partners need payment terms, reporting requirements and workload expectations they can sustain.
A public-health coalition should therefore have more than a shared aspiration. It needs a population definition, a service-capacity commitment, a funding arrangement, a route to resolve cross-agency failures and a common review of who is still not being served.
Evaluate the pathway and the endpoint
Track screened, eligible, reached, accepted, served and followed up as separate denominators. Then assess the intended outcome at the appropriate horizon. Improved engagement is useful evidence, but it does not by itself establish better health or lower total spending.
Before attributing a change to the program, examine baseline differences, regression to the mean, changing enrollment, incomplete claims and concurrent services. A fall in spending among people selected for unusually high prior spending is not enough to establish impact. Where randomization is impractical, the comparison design and its limitations still need to be explicit.
AI can help identify patterns, reconcile records and prioritize review. It should not turn incomplete historical utilization into a definitive judgment about a person’s need. Clinical review, patient preferences and a route to challenge or update the assessment belong in the allocation design.
The public-health leadership agenda
I would bring five decisions to the leadership table: the population we owe accountability to; the outcomes we will protect; the constraint we will fund first; the institutions that must share the commitment; and the evidence that will change the next allocation.
The resulting portfolio should explain both its intensive interventions and its universal foundations. Targeted support and broad access infrastructure are complementary choices. The measure of strategic quality is whether scarce resources produce a defensible improvement in population well-being—not whether the dashboard contains more risk scores.
Sources & analytical basis
The strategic recommendations are the author’s interpretation. External research and company observations are attributed below; illustrative scenarios are labeled where used.