From claims reporting to decision intelligence: what healthcare funders need next

Ask a healthcare funder for its claims data and you will usually receive a great deal of it. Monthly claims reports, utilisation summaries, loss-ratio trends, provider league tables, high-cost claimant lists. The reporting function is rarely the constraint. The constraint is that all of this describes what happened without telling anyone what to do about it.

This is the difference between claims reporting and decision intelligence. Reporting answers "what happened?" Decision intelligence answers "what should we change, and what will happen if we do?" Most funders have built the first and assume the second will follow from it. It does not follow automatically, and the gap between them is where a lot of value is lost.

Why more reports do not close the gap

Reporting is organised around data sources — claims, membership, providers — because that is how the systems that produce it are organised. Decisions are organised around choices: which benefit to change, which network to tighten, which product to reprice, which programme to fund. The two are not the same shape, and no amount of additional reporting reshapes one into the other.

A claims report can show that a benefit line is running hot. It cannot tell you whether the driver is price, volume, case mix, provider behaviour or a design flaw in the benefit itself — and those five causes call for five different responses. The report ends exactly where the decision begins. Adding a sixth report does not help; it adds another description of the same symptom.

Decision intelligence starts from the other end. It begins with the decision on the table and works backward to the specific evidence that would change it. That reframing is what turns a data asset into a management asset.

The layers between data and decision

Moving from reporting to intelligence is a progression, and it is useful to be honest about which layer an organisation is actually operating at.

Description. What happened — claims, utilisation, loss ratios. Necessary, and where most funders are strong.

Diagnosis. Why it happened — separating price from volume, case mix from provider behaviour, genuine risk from leakage. This is where reporting usually stops and where the first real intelligence begins.

Anticipation. What is likely to happen next — scenario and sensitivity testing, projecting the effect of a benefit change or a network shift before committing to it. This is where analysis starts to shape decisions rather than explain them.

Decision. What to do — the point where the analysis resolves into a specific, evidenced recommendation a board can act on.

The value is concentrated in the last two layers, but most investment goes into the first. Better dashboards make the description layer prettier. They do not, on their own, move an organisation up the ladder.

What decision intelligence looks like in practice

Consider a funder facing rising day-to-day claims. The reporting layer confirms the rise and quantifies it. Decision intelligence takes it further. It isolates whether the increase is concentrated in specific providers, specific member segments or specific benefit lines. It tests whether the pattern reflects genuine need, a design incentive that rewards over-use, or leakage that controls should be catching. It models what a targeted benefit or network change would do to both cost and member value. And it presents the trade-off as a decision, with the evidence attached.

The output is not a longer report. It is a short, defensible recommendation — this is the driver, this is the option, this is the expected effect, this is the risk — that an executive can take into a board meeting and stand behind.

Where advanced techniques fit

Predictive modelling and machine learning belong in this discussion, but they belong at the right layer and under the right conditions. Advanced statistical and predictive techniques are valuable where the data quality, the governance and the mandate justify them — for example in anticipating high-cost claimants or detecting patterns of fraud, waste and abuse that rules-based checks miss.

They are not a substitute for the diagnosis layer, and they are not a reason to skip it. A predictive model built on poorly understood claims data will produce confident answers to the wrong question. The discipline is to apply sophistication where it earns its place, and not to reach for it as a way around the harder work of understanding the claims.

What this asks of funders

The move from reporting to decision intelligence is less a technology programme than a change in how analysis is commissioned and consumed. It asks three things.

First, start from the decision. Before building another view, name the choice it is meant to inform. Analysis without a decision attached is reporting by another name.

Second, invest in diagnosis, not just description. The capability to separate cause from symptom is where reporting becomes intelligence, and it is usually under-resourced relative to its value.

Third, hold analysis to a decision standard. The test of a piece of work is not whether it is accurate — it should be — but whether an executive can act on it. Evidence that does not change a decision has not done its job.

Funders do not need more data. They need the distance between the data and the decision closed. That is a design choice about how intelligence is built and used, and it is available to any funder willing to make it.