The Med-Mal Funnel: Most Claims Pay Nothing

TL;DR — Stop allocating top-tier defense counsel to claims destined to close at zero. By separating liability probability from severity potential on day one, claims leaders can triage files aggressively and stop wasting defense spend on phantom threats.

A plaintiff attorney files a broad complaint naming every doctor, nurse, and hospital in the zip code. The claims desk receives the file, accompanied by thousands of pages of medical records. The natural instinct is to brace for impact. Nuclear verdicts dominate industry headlines, and social inflation drives a narrative of perpetual severe losses. But the statistical reality of the medical malpractice claim funnel tells a completely different story. The vast majority of these claims will pay absolutely nothing in indemnity. They will be dropped when the plaintiff attorney realizes the medical records do not support a breach of the standard of care. They will be dismissed on summary judgment. Or the defense will win an outright victory at trial.

The indemnity payout on these files is zero. The defense cost is staggering. Carriers bleed capital paying hourly rates to investigate, depose, and defend files that never had a viable path to a plaintiff victory. Claims professionals treat every new file as a potential policy-limit loss because they lack the tools to separate liability probability from severity potential early in the lifecycle. When a file arrives, the medical records are dense and unstructured. It takes months of expensive defense counsel time just to figure out what actually happened in the operating room. By the time the carrier realizes the claim is meritless, the defense budget is already blown. The funnel is working exactly as designed for the plaintiff bar, forcing the defense to spend heavily just to prove a negative.

The Trap of Blended Reserving

This uncertainty breaks the reserving process on day one. When a claims handler cannot determine if a case is a zero-payout dismissal or a severe settlement, they often split the difference. They set a blended reserve. This is a mathematical fiction. A medical malpractice claim does not resolve for the average of zero and ten million dollars. It resolves for zero, or it resolves for ten million. A blended reserve guarantees the carrier is wrong from the start. It locks up unnecessary capital on meritless claims and leaves the book dangerously under-reserved on the actual threats. As the case drags on, claims teams engage in step-stair reserving, slowly bumping up the number as bad news trickles in from defense counsel.

We see this dynamic clearly through the lens of the Priest-Klein hypothesis. Cases where liability is clear tend to settle quickly. Cases where liability is highly uncertain, or heavily skewed toward the defense, proceed through litigation. The cases that reach a jury are not representative of the whole funnel. They are the anomalies where the parties could not agree on the probability of a defense verdict. If claims teams rely on the headlines of nuclear verdicts to set their internal priors, they warp their view of the entire funnel. They begin applying verdict-level fear to settlement-level realities. You must keep verdict-heavy comparable cases in a separate analytical module from settlement-anchored data. Mixing them creates a distorted picture that drives up settlement values unnecessarily.

The operating model fails because it relies on human triage of unreadable volumes of data. You cannot negotiate from data or allocate defense spend rationally if you wait for defense counsel to summarize the medical files six months after notice. The unstructured data inside those files holds the answer. The surgical notes, the expert correspondence, and the plaintiff pleadings contain the specific drivers of liability. But a human claims adjuster handling a full desk cannot read four thousand pages of medical history in an afternoon. They default to assigning standard defense counsel and waiting for a report.

Separating Probability from Severity

Fixing this requires a fundamental shift in how claims organizations process information at the top of the funnel. Generative AI is built to read and structure this massive volume of text. It does not predict the outcome. It simply extracts the facts, the timelines, the pre-existing conditions, and the specific drivers of the claim. It does the reading and the structuring. Once that text is structured into a usable format, separate mathematical and geometric machine-learning models take over. These models compare the geometry of the current claim against a large volume of resolved cases with known outcomes.

The output is not a single point guess. It is a calibrated settlement range, an escalation probability, and a clear view of comparable resolved cases. It provides a reserve delta versus the current reserve, entirely traceable to the source documents. By isolating the probability of liability from the potential severity of the injury, claims leaders can make decisive choices. They can assign standard counsel to the highly probable zero-payout files, holding the line on defense costs. For the files showing high liability probability and severe injury characteristics, they can allocate top-tier counsel, set a realistic reserve on day one, and initiate early settlement discussions.

Early detection is the only defense against third-party litigation funding. When a file sits dormant while defense counsel reviews medicals, litigation funders assess the risk and inject capital into the plaintiff's war chest. Late escalation guarantees you are negotiating against a well-funded adversary who has already anchored their expectations to a nuclear outcome. When you have a calibrated settlement range on day one, you dictate the timeline. You approach the negotiation with honest uncertainty bands and field-specific priors, removing the emotional leverage the plaintiff relies on.

The medical malpractice funnel will always be wide at the top. The goal is not to stop the initial filings. The goal is to stop funding the illusion that every filing is an existential threat. When you can read the reality of a file on day one, you stop paying defense counsel to discover it for you.