Paid medical malpractice claims: How strongly does the past predict the future?

• Published date: 01 December 2023
• Author: Kowsar Yousefi,Bernard Black,David A. Hyman
• Date: 01 December 2023
• DOI: http://doi.org/10.1111/jels.12371

ORIGINAL ARTICLE

Paid medical malpractice claims: How strongly does the

past predict the future?

Kowsar Yousefi

1,2

| Bernard Black

3

| David A. Hyman

4

1

Institute for Management and Planning Studies, Tehran, Iran

2

Sharif University of Technology, Tehran, Iran

3

Northwestern University, Evanston, Illinois, USA

4

Georgetown University Law Center, Washington, DC, USA

Correspondence

David A. Hyman, Scott K. Ginsburg Professor of Health Law & Policy, Georgetown University Law Center,

Washington, DC, USA.

Email: dah137@georgetown.edu

Abstract

When does the past predict the future? In financial markets, warnings that “past results

are no guarantee of future performance”are ubiquitous. But in multiple fields (including

professional sports, insurance, and criminal law), it is widely believed that the past is a

useful guide to the future. Does that insight apply to medical malpractice (“med mal”)?

Using a novel dataset (which includes detailed data on all licensed physicians and all paid

claims in Illinois over a 25-year period), we study whether past paid med mal claims, phy-

sician characteristics, and specialty predict future paid med mal claims. After controlling

for other factors, physicians with a single prior paid claim have a fourfold higher risk of

future claims than physicians with zero prior paid claims. The more prior paid claims a

physician has, the higher the likelihood of a future paid claim. Multiple factors (male

gender, having an MD, attending a non-U.S. medical school, practicing in a high-

claim-risk specialty, and mid-career status [6–15 prior years of experience]) predict a

higher likelihood of having one or more paid med mal claims.

KEYWORDS

medical malpractice, survival analysis

INTRODUCTION

When does the past predict the future? In financial markets, warnings that “past

results are no guarantee of future performance”are ubiquitous. But in fields

ranging from professional sports to insurance to criminal law, it is widely

DOI: 10.1111/jels.12371

©2023 Cornell Law School and Wiley Periodicals LLC.

818 J Empir Leg Stud. 2023;20:818–851.

wileyonlinelibrary.com/journal/jels

believed that the past is a useful guide to the future, and an entire literature is

devoted to each of these topics. Although the names of the fields may differ

(e.g., sabermetrics, actuarial sciences, recidivism, the “hot hand”), the underly-

ing issue remains the same.

Within tort law, researchers have focused on medical malpractice—and

whether physicians with one or more prior paid claims are more likely to have

future paid claims. We contribute to this literature by using a novel dataset,

which contains detailed personal information (including paid claim history) for

the 84,114 physicians with an active license in Illinois at any time over a 25-year

period (1990–2014). We apply multiple failure survival analysis—principally the

nonparametric Kaplan–Meier survival model and the Cox semi-parametric pro-

portional hazard model. However, our results are robust to alternative models,

including Gompertz and Weibull parametric multiple failure models and

Kaplan–Meier and Cox single failure models (all of which we present in the

Appendix).

Most prior studies of the predictive power of prior paid claims have focused

on the minority of physicians—in many specialties, a small minority—who have

at least one paid claim and measured the risk of a future claim for a physician

who already has one or more prior claims. Because we also have data on the

92% of physicians in Illinois with zero paid claims, we have a richer dataset with

which to analyze the factors that predict paid claim risk and can compare the

risk of a future claim for physicians with 1 prior claim to the risk for physicians

with 0 prior claims.

We find that physicians with 1 prior paid claim have an annual risk of future

paid claims that is roughly fourfold higher than physicians with 0 prior claims.

The increase in future claim risk is sixfold for physicians with 2 prior paid claims

and eightfold for physicians with 3+prior paid claims.

We also study the predictive effect of other physician characteristics.

Overall, the hazard ratio for a paid claim for female physicians is about half that

of male physicians. Physicians with an MD degree have a higher paid claim risk

than those with a DO degree. Physicians who attended a non-U.S. medical

school have higher paid claim rates than graduates of U.S. medical schools.

Unsurprisingly, overall paid claim risk is higher in high-claim-risk specialties

(i.e., surgery, ob-gyn). However, our results for other predictive factors are simi-

lar across both high-claim-risk and low-claim-risk specialties.

LITERATURE REVIEW

As noted above, there is substantial literature on the association between past

med mal claims and future med mal claims. The topic is important because it

bears on the value of paid med mal claims as a signal of physician quality and

on the basis for state laws that seek to limit med mal lawsuits. Those laws are

PREDICTING FUTURE MALPRACTICE 819

often premised on the view, held by many physicians, that med mal claims are

largely random events unrelated to quality of care. These studies typically ana-

lyze the association between past and future claims, as well as the impact of var-

ious demographic covariates (e.g., specialty, gender, years of experience, etc.).

For ease of explication, we divide our summary into studies that focus on the

predictive value of prior claims, and studies that focus on the impact of various

covariates (e.g., gender, specialty, degree type, and location of medical school)

on claiming.

Predictive value of prior claims

Bovbjerg and Petronis (1994) use Florida data from 1975 to 1988, for physicians

with and without prior paid claims. They define a baseline period (1975–1980),

and a subsequent period (1981–1988). They report that less than 7% of physi-

cians had any claim (paid or unpaid) during the subsequent period, but that

physicians with 1+claims in the baseline period had a 2.84 times higher risk for

a paid claim during the subsequent period. Having a single unpaid baseline

claim roughly doubled the odds of a subsequent-period claim.

Sloan et al. (1989) also use Florida data, from 1975 to 1983 for physicians

with and without prior paid claims. They use 1975–1980 as the baseline period,

and 1981–1983 as the subsequent period. Physicians with no paid claims during

the prior period had a 15% chance of having a paid claim in the future period,

compared to 36% for physicians with prior paid claims. Thus, having a prior

paid claim more than doubled the risk of a future claim.

These two studies are the only prior studies that included physicians without

prior paid med mal claims.

Studdert et al. (2016) use the National Practitioner Data Bank (NPDB), a

national repository of paid claims against physicians. Their study (and all

other studies using the NPDB) have data only on physicians with paid

claims. Using an Andersen–Gill, multiple failure survival model, similar to

ours, they assess the extent to which physicians with two or more prior paid

claims face a higher risk of a future paid claim compared to physicians with

one prior paid claim. They find that “the risk of recurrence increased with

the number of previous paid claims.”For example, compared to physicians

with a single-paid claim, physicians with three paid claims had three times

the risk of a future claim. Hyman et al. (2023)andBlacketal.(2019)also

use the NPDB and report that having even a single paid claim in an initial

5-year baseline period roughly quadruples the risk of having a paid claim in

the next 5 years.

These three studies all report that prior paid claims predict future paid

claims. However, Studdert et al. do not study the 94% of physicians in their

sample with no paid claims, and Hyman et al. and Black et al. study these

820 PREDICTING FUTURE MALPRACTICE