Defense counsel walk into mediation with a number. It came from a reserve set months or years ago, adjusted by whatever has happened since, sanity-checked against a handful of cases the firm has handled before.

That number is usually the single most consequential input in the room. It is also, in our experience reading the historical data, the least examined.

The mistakes below are not failures of lawyering. They are failures of benchmarking, and they share a common shape: a defensible-sounding number, produced by a process nobody stress-tested, treated as ground truth for the rest of the negotiation.

1. Anchoring to a reserve that was set against stale severity

The most common error is also the least visible, because the reserve feels like an objective starting point rather than a prediction.

Medical malpractice severity has been rising, and recent-year data understates how much. As S&P Global Market Intelligence has noted, the apparent "moderation in newer years" in severity figures is largely an artifact of claim immaturity: malpractice claims take years to resolve, so recent accident years show only the early, pre-resolution slice of a claim population that will keep moving upward.

The practical consequence for defense counsel is uncomfortable. A reserve benchmarked against 2022 or 2023 accident-year severity was set against incomplete data, and the direction of the error is predictable. It is too low. Not sometimes, systematically.

What to do instead: treat the reserve as a hypothesis with a date attached. Ask what data window produced it and whether that window has matured. A reserve set three years ago against then-current comparables is not a conservative number. It is an old number.

2. Negotiating from a point estimate when the real exposure is in the tail

Reserves are single numbers. Exposure is a distribution. A mediation strategy built on the first while ignoring the second gets blindsided in a specific, repeatable way.

Two cases can share an identical expected value and carry completely different risk. One clusters tightly around the median. The other has a long right tail, where a meaningful share of comparable cases resolved far above the expected value. The expected value tells you nothing about which case you are sitting in.

This matters most in exactly the situation where defense counsel feel most confident: the case that "should" resolve modestly. If the severity distribution for that case profile is wide, the modest expected value is an average of outcomes that includes results your client cannot absorb.

What to do instead: ask for the spread, not just the midpoint. The question that should drive settlement authority is not "what is this case worth?" but "what share of cases like this one resolved above the number I am authorized to pay?"

3. Applying a national reserve policy to a local venue

Jurisdiction produces the strongest single effect in the malpractice data, and it is not close.

Median payout in California: $73,000. Median payout in Pennsylvania: $507,000. Comparable case characteristics, roughly seven times the outcome. California's MICRA cap has historically compressed payouts; Pennsylvania has no comparable cap and includes plaintiff-friendly venues in Philadelphia.

Carriers and national firms tend to set reserve philosophy centrally and apply it broadly. That is administratively sensible and analytically indefensible. A national reserving standard is systematically too high in capped states and too low in uncapped ones, which means it is wrong nearly everywhere, just in opposite directions.

Venue also operates below the state level. Urban counties in high-population states produce different verdicts than rural counties in the same state. A Philadelphia case and a rural Pennsylvania case are not the same case.

What to do instead: anchor to the venue, then adjust for case facts. Not the reverse.

4. Treating death as the ceiling of exposure

This one is genuinely counterintuitive, and it runs directly against the instinct most defense attorneys have developed.

Median payout for cases resulting in death: $311,000. Median payout for cases resulting in grave permanent injury: $807,000.

Death is not the ceiling. Grave permanent injury cases typically involve ongoing care costs, lost future earning capacity, and a plaintiff who can appear and testify about their own suffering. Wrongful death damages are often capped or structurally limited in ways that severe-injury damages are not.

The defense-side implication is direct: the cases most likely to be under-reserved are surviving catastrophic-injury cases, precisely because they do not trigger the alarm that a death case does. If your reserving instinct treats a death as maximum exposure and a severe permanent injury as something less, that instinct is inverted relative to the data.

What to do instead: reserve severe-injury survival cases at least as aggressively as comparable death cases, and expect plaintiff's counsel to understand this asymmetry even when your own file does not reflect it.

5. Treating practitioner history as atmosphere rather than a number

Most defense files note prior claims history somewhere. Few price it.

Practitioners with five or more prior adverse reports settle for 43% more than first-time reporters. That is not a soft factor or a jury-appeal concern. It is a quantifiable multiplier that shows up consistently in the historical record.

Prior history tends to get discussed qualitatively in defense strategy sessions: it is "bad facts," something to keep out if possible, a reason the case feels uncomfortable. Then the reserve gets set as though it were a first-time case with an asterisk.

What to do instead: research the history, then apply it to the number rather than to the mood of the meeting. If the multiplier is real and you have not priced it in, opposing counsel has, and they will hold at a number you were not prepared for.

The pattern underneath all five

Every mistake above is the same mistake wearing different clothes. Each one substitutes a familiar reference point for a relevant one: last year's comparable, the firm's prior experience, the national policy, the intuitive severity ranking, the qualitative read on the practitioner.

Anchoring is not a defense-side failing specifically. It is how experienced professionals compress large amounts of experience into fast judgments, and most of the time it works. It fails in med mal because the variance across jurisdiction, injury severity, allegation type, and practitioner history is large enough that the familiar reference point is frequently the wrong one, and nothing in the negotiation surfaces the error until the number is already on the table.

The remedy is not to distrust judgment. It is to check the anchor before the mediation rather than after it.

Bringing data into the room

MedMalPredict produces case-specific predictions from more than 270,000 historical malpractice records: payment probability, a payout range rather than a single figure, and the distribution of outcome severity across comparable cases.

For defense counsel, the third output is often the most useful and the most overlooked. Knowing that a case type resolves with a wide severity spread changes settlement posture in a way that no expected value can, and it converts "this feels risky" into a number you can bring to a client who wants to hold firm.

The prediction does not make the call. It tells you whether the number you walked in with can survive contact with the historical record.

Try It

If you have a mediation on the calendar, run the case profile before you set your authority.

Try MedMalPredict

MedMalPredict AI is not legal advice. Predictions are based on historical data and represent probabilities, not guarantees.

What Defense Attorneys Get Wrong at Mediation