The Conversations That Prevent Claims: What the Research Actually Shows

By John Mahoney · August 2026 · 12 min read

Start with the fact that makes the rest of this coherent: the relationship between medical injury and malpractice claims is loose in both directions.

The Harvard Medical Practice Study matched the records of more than 31,000 hospitalized patients in New York against the state's malpractice claims data. Of the 280 patients who suffered an adverse event caused by negligence, eight filed a claim — a weighted rate of 1.53% (Localio et al., NEJM 1991;325:245–51). Most people genuinely harmed by negligence never sue.

Now the other direction. When Studdert and colleagues had physician reviewers examine 1,452 closed claims, they found that 37% did not involve an identifiable error, though most such claims were denied compensation, while 73% of claims involving injury from error were paid (Studdert et al., NEJM 2006;354:2024–33). The system sorts reasonably well in the end — slowly, and at 54 cents of administrative cost per dollar of compensation.

Put those together and one conclusion is unavoidable. Whether an injured patient files is only partly a function of what happened to them clinically. Something else is doing a lot of the work — and that something has been studied for thirty years. Here is what that literature actually found, including the parts that cut against the tidy version.

Why patients say they sue

The first serious attempt to ask them directly was Vincent, Young, and Phillips, who surveyed 227 patients and relatives pursuing legal action through five firms of plaintiff medical negligence solicitors (Lancet 1994;343:1609–13). Four motivations dominated: the need for an explanation; a desire to ensure the same thing would not happen to someone else; accountability; and compensation. The decision to litigate was determined not only by the original injury but by how the aftermath was handled — fewer than 15% of the explanations patients received were considered satisfactory. Among respondents who thought something could have prevented the suit, roughly 40% said an explanation and an apology would have removed the need to file (as summarized in Robbennolt, Clin Orthop Relat Res 2009;467:376–82).

Hickson and colleagues interviewed 127 mothers of infants with permanent injury or death whose Florida claims had closed (JAMA 1992;267:1359–63). Among reasons volunteered for filing: 33% had been advised to by a knowledgeable acquaintance, 24% cited recognizing a cover-up, 20% needed information. The communication findings are worth sitting with: 32% said their physicians would not talk openly, 48% believed they had been misled, and 70% said no one warned them about long-term neurodevelopmental problems.

Beckman, Markakis, Suchman, and Frankel coded 45 plaintiff depositions — 3,787 pages of transcript (Arch Intern Med 1994;154:1365–70). Problematic relationship issues appeared in 71% of them, in four themes: deserting the patient (32%), devaluing patient or family views (29%), delivering information poorly (26%), and failing to understand the patient or family perspective (13%). Single institution, small sample — but the themes have proven durable.

Read together, these describe the same person: someone who was hurt, could not get a straight answer, felt dismissed when they asked, and eventually found the one institution in American life obligated to answer questions under oath.

The communication research, reported honestly

The most cited study in this field is Levinson and colleagues, JAMA 1997;277:553–9. Researchers audiotaped ten routine office visits for each of 59 primary care physicians and 65 general and orthopedic surgeons, then compared the communication behavior of physicians with prior claims against those without.

Among primary care physicians, the no-claims group differed in specific, teachable ways. They used more statements of orientation — telling patients what to expect and how the visit would go. They laughed and used humor more. They tended to use more facilitation: soliciting opinions, checking understanding, encouraging the patient to keep talking. And their visits ran longer — a mean of 18.3 versus 15.0 minutes (abstract).

Here is the part that usually gets dropped in the retelling, and it matters: none of those differences were found among the surgeons. The authors state it plainly — significant differences in communication behavior between no-claims and claims physicians were identified in primary care physicians but not in surgeons, and multivariable models did not improve prediction of claims status for surgeons at all.

The honest summary is not "be nicer and you will not be sued." It is narrower:

In longitudinal primary care relationships, communication behavior is measurably associated with claims history. In procedural specialties, this method did not detect that association — which may reflect a different relationship structure, a different injury profile, or the limits of audiotaping ten routine visits. These are associations, not proof of causation.

Keep that in mind whenever this literature is quoted at you as a scheduling intervention. Longer visits are associated with fewer claims; adding three minutes to a template is not a liability program. The behaviors underneath — orienting, facilitating, checking understanding — are the plausible active ingredients, and they take less time than they appear to.

What happened when institutions changed the aftermath

If the aftermath drives filing, changing the aftermath should change filing. Health systems and carriers tested that with what are now called communication-and-resolution programs (CRPs): investigate quickly, tell the patient what was found, apologize where an apology is owed, and proactively offer compensation when the care was unreasonable.

The University of Michigan disclosure-with-offer program

The most-cited outcome study is Kachalia et al., Ann Intern Med 2010;153:213–21, spanning 1995–2007. After full implementation:

Average monthly rate of new claims fell from 7.03 to 4.52 per 100,000 patient encounters.
Average monthly rate of lawsuits fell from 2.13 to 0.75 per 100,000 patient encounters.
Median time from claim report to resolution fell from 1.36 to 0.95 years, with substantial reductions in legal defense costs.

The authors were careful with it, and you should be too: this is a before-after design in a single academic system with a captive insurer, over a period when claim rates were falling in many places. Strong supporting evidence, not a randomized trial.

A carrier that built it into the policy

COPIC, a Colorado physician-owned liability insurer, has run its 3Rs program — recognize, respond, resolve — since 2000. Insured physicians disclose early and stay in the relationship; the carrier reimburses out-of-pocket medical expenses (up to $25,000) and loss of time (up to $5,000), with no release of claims and no admission of liability. One caution: the striking outcome figures widely attributed to 3Rs circulate without a traceable primary source, so treat any percentage quoted about it skeptically. The structural point survives without the numbers — a malpractice carrier built early disclosure into its risk-management product, which is not what you would expect if silence were optimal.

The federal toolkit

The Agency for Healthcare Research and Quality publishes the CANDOR toolkit (Communication and Optimal Resolution), an eight-module package covering event investigation, disclosure communication, resolution, care for the caregiver, and organizational learning. It is free, it is the closest thing to a national standard for this work, and the care for the caregiver module is the one most physicians do not know exists. Be precise about its status, though: CANDOR is a well-documented toolkit, not an intervention with a published outcome evaluation of its own.

And the results that were mixed

Mello and colleagues evaluated the Massachusetts CARe program (Communication, Apology, and Resolution) across four hospitals and reported reason for optimism in 2017: only 5% of events entering the program led to a claim or lawsuit, and the minority meeting compensation criteria resolved at a median payment of $75,000.

Then Kachalia, Mello, and colleagues examined liability trends across four Massachusetts hospitals against non-implementing peers (Health Aff 2018;37:1836–44). Mixed: improved trends in new-claim rates and defense costs at some hospitals, no significant change in other outcomes, and no hospital where liability trends worsened. That is a "no harm" finding, not a "big savings" finding, and it is materially weaker than the Michigan result.

And in five New York City hospitals implementing a CRP in general surgery, all five improved disclosure and adverse-event surveillance but none could fully implement the compensation component (Mello et al., Health Aff 2014;33:30–38). Programs on paper do not produce the results of programs actually run.

The fair conclusion: CRPs do not increase liability exposure, they shorten resolution, and their effect on claim rates ranges from substantial to null depending on implementation. The gap between "communication matters to patients" — very well supported — and "communication programs reduce liability" — supported unevenly — is the real state of this literature.

Apology laws: know what they do and do not do

By the National Conference of State Legislatures' tally, thirty-nine states plus the District of Columbia have some form of apology statute making certain post-adverse-event statements inadmissible. (That tally was last updated in 2021 — check your own state's current statute rather than any national count, this one included.) They fall into two categories, and the difference is the entire ballgame:

Narrow (sympathy-only) statutes — the large majority, roughly 34 states plus D.C. — exclude expressions of sympathy, condolence, regret, or compassion. "I am so sorry this happened to you" is protected. "I made a mistake" is not; admissions of fault remain admissible.
Broad (fault-inclusive) statutes extend the exclusion to admissions of fault or error made in connection with an expression of sympathy. Only a handful of states have them — Arizona, Colorado, Connecticut, Georgia, and South Carolina. Georgia's is the clearest illustration, reaching statements expressing regret, apology, sympathy, commiseration, condolence, compassion, mistake, or error.

Our attorney-side survey of apology statutes by state walks the categories in detail. What a clinician needs is simpler: you almost certainly do not know which category your state is in, and the answer changes what is safe to say. Ask before you need to know.

You should also know that the statutes have not performed as advertised. McMichael, Van Horn, and Viscusi analyzed claims data from a large national insurer covering 2004–2011 and found that apology laws did not reduce liability risk ("Sorry" Is Never Enough, Stanford Law Review 71:341 (2019)). For surgeons, no significant effect on any litigation outcome. For non-surgeons the direction ran the wrong way: the probability of facing a lawsuit rose by about 1.2 percentage points — roughly a 46% increase relative to the national average — and indemnity payments more than doubled. The proposed mechanism is information: where a patient could not otherwise tell something went wrong, an apology is itself a signal.

Two caveats belong with that finding. It measures statutes, not apologies — it cannot tell you whether any individual apology helped or hurt. And the estimates are for the narrow, sympathy-only laws that make up the large majority, which may be the actual lesson: a shield covering "I am sorry" but not "here is what happened" may be the worst of both worlds.

The lesson is not that apologizing is dangerous. It is that a statute is not a program. An apology law protects a sentence; it does not investigate the event, answer the family's questions, or offer compensation. The institutional programs that changed outcomes did all three. A physician apologizing into a vacuum, with no explanation to follow and no institutional resolution behind it, is offering the one thing patients said they wanted least on its own.

What to say after an adverse outcome

Your carrier's guidance governs. But the shape of the conversation is well established, and not complicated.

Do:

Go back in the room. Deserting the patient was the largest theme in the deposition analysis. Presence after a bad outcome has the best evidence behind it and the lowest cost.
Express sympathy without conditions. "I am so sorry this happened. This is not what any of us wanted for you." Protected in most states, and owed regardless.
Say what is known and what is not. "Here is what we know. Here is what we do not know yet. Here is how we will find out, and when I will come back and tell you." Then keep that appointment — the broken promise to follow up is its own injury.
Answer the question that was asked. Patients overwhelmingly report wanting an explanation; withholding one to avoid liability is the behavior most reliably associated with claims in this literature.
Say what will change. Preventing recurrence was one of the four motivations patients named. If a review is underway, say so.
Notify risk management and your carrier the same day — not to build a defense, but because they run the disclosure process and bring resources you cannot.

Don't:

Speculate about causation before the facts are in. "The alarm must have been missed" is a guess that names someone else and may be wrong.
Assign fault on incomplete information — yours or anyone's. That differs from acknowledging what happened: "the drain was placed on the wrong side" is a fact; "I was negligent" is a legal conclusion.
Blame the system to the family. It is usually heard as deflection.
Go dark. Silence is the pattern in every one of these studies.
Alter the record afterward. The chart entry about the conversation must be contemporaneous and factual — see documentation that prevents lawsuits — and stop entirely once a claim is on the horizon.

If the conversation did not prevent it, preparation is the next lever.

Survive Your Deposition is 15 modules built for physicians — each one ends with a live AI drill where you answer out loud and get scored on what you actually said, not on what you meant.

See the course →

The bottom line

The evidence does not show that good communication makes you lawsuit-proof; the surgical null result in Levinson and the mixed Massachusetts outcomes are both real, and anyone selling certainty is overselling. What it does support is narrower and still worth a great deal: patients who are told the truth and treated as people entitled to answers behave differently than patients who are not. Institutions that formalized that saw claims resolve faster and, in the strongest studies, less often.

"Never apologize" is now clearly behind the evidence — carriers themselves build disclosure programs, and a federal agency publishes the implementation toolkit. The modern version of the advice is not say less. It is say the true thing, early, with your institution beside you, and know which category of apology statute your state has before you need to.