Body of Evidence: Do autopsy findings impact medical malpractice claims outcomes?

Abstract

Objective

Clinicians may hesitate to advocate for autopsies out of concern for increased malpractice risk if the pathological findings at time of death differ from the clinical findings. We aimed to understand the impact of autopsy findings on malpractice claim outcomes.

Methods

Closed malpractice claims with loss dates between 1995–2015 involving death related to inpatient care at three Harvard Medical School hospitals were extracted from a captive malpractice insurer’s database. These claims were linked to patients’ electronic health records (EHRs) and their autopsy reports. Using the Goldman classification system, two physician reviewers blinded to claim outcome determined whether there was major, minor or no discordance between the final clinical diagnoses and pathologic diagnoses. Claims were compared depending on whether an autopsy was performed and whether there was major versus minor/no clinical-pathologic discordance. Primary outcomes included percentage of claims paid through settlement or plaintiff verdict and the amount of indemnity paid, inflation-adjusted.

Results

Of 293 malpractice claims related to an inpatient death that could be linked to patients’ EHRs, 89 claims (30%) had an autopsy performed by either the hospital or medical examiner. The most common claim allegation was an issue with clinician diagnosis, which was statistically less common in the autopsy group (18% vs 38%, P=0.001). There was no difference in percentage of claims paid whether an autopsy was performed or not (42% vs 41%, P=0.90), and no difference in median indemnity of paid claims after adjusting for number of defendants ($1,180,537 vs $906,518, P=0.15). Thirty-one percent of claims with hospital autopsies performed demonstrated major discordance between autopsy and clinical findings. Claims with major clinical-pathologic discordance also did not have a statistically significant difference in percentage paid (44% vs 41%, P>0.99) or amount paid ($895,954 vs $1,494,120, P=0.10) compared to claims with minor or no discordance.

Conclusions

While multiple factors determine malpractice claim outcome, in this cohort, claims in which an autopsy was performed did not result in more paid outcomes, even when there was major discordance between clinical and pathologic diagnoses.

Introduction

The Joint Commission previously required hospitals to conduct autopsies on a minimum percentage of inpatient deaths. The threshold was 25% for teaching hospitals and 20% for community hospitals.¹ After this requirement was rescinded in 1971, the autopsy rate in United States hospitals declined from 40–60% of all inpatient deaths to less than 5% in the mid 1990s.^2,3 The reasons for global decline in the rate of autopsies are multifactorial, and include increased emphasis on reducing education and healthcare costs, advances in imaging and other diagnostic modalities, cultural aversion, and modern pathology training and practice structure.⁴

Nevertheless, autopsy remains a critical element of health care quality control and results from autopsies can provide important information for patients’ families and clinicians. Autopsies may provide closure for family members when there is clinical uncertainty, uncover important genetic implications for next of kin, and even help control disease outbreak.⁴ They can be used in medical education to teach the pathologic basis of disease⁵ and drive quality improvement when there is a missed diagnosis, thus promoting transparency, learning, and accountability among the whole medical team.^6,7 For this reason, however, clinicians may worry about increased malpractice risk if the autopsy findings suggest a diagnosis that is different from the clinical diagnosis at the time of death,⁴ particularly if that diagnosis was potentially actionable. Previous studies have demonstrated that new findings are identified in approximately 50% of autopsies⁸ and diagnostic errors revealed in about 30% of autopsies.⁹

Previous studies examining autopsies and medical malpractice outcomes have reported mixed results and have focused on the pediatric population.^10,11 Given the potential for autopsies to provide value to clinicians, health systems and patients’ families, we aimed to better understand the impact of autopsies on the outcome of malpractice claims, especially when there is discordance between the clinical and pathologic diagnoses.

Methods

Study Design and Claim Extraction

We performed a retrospective cohort study of closed malpractice claims with a loss date between 1995–2015 that resulted in either death as an inpatient or death related to inpatient care at three Harvard Medical School affiliated hospitals in Massachusetts. A claim is defined as a written demand for compensation¹² and loss date is defined as the date on which the harm occurred, not the date on which the harm was identified. These hospitals are insured by the Controlled Risk Insurance Company (CRICO), a captive malpractice insurer. Two of the hospitals are tertiary academic centers and one is a community hospital.

This study was approved by the Partners HealthCare’s Institutional Review Board (2015P002516).

Data Source

At the time of liability claim, physicians at CRICO review the clinical aspects of the individual cases. Nurses, who are full-time taxonomy specialists, code the claims using a proprietary clinical coding taxonomy. The taxonomy, which captures both systems and clinical factors, and is applied to a larger national database cited in previous articles^13–15 and is described in detail in the eMethods. The coding undergoes robust quality assurance, including regular coder conference calls, annual coders’ conferences and auditing processes where 15% of claims are randomly selected and reviewed quarterly to ensure consistency and inter-rater reliability.

Data Collection

We linked all claims involving an inpatient death with loss date between 1995–2015 (n=293 claims) to their respective electronic health records (EHRs) using the patient’s medical record number. The mother’s medical record number was used for perinatal cases. Blinded to outcome, two physician reviewers (LCM and RMG) from different specialties (internal medicine and surgery, respectively) examined the two sources of data, the malpractice claim file and the EHR, as follows.

First, the malpractice claim file and EHR were reviewed to determine if an autopsy had been performed, including if there was documentation of the Medical Examiner performing an autopsy outside of the hospital. For perinatal patients to be considered as having an autopsy, bodily tissues had to be examined, not just the placental tissue that is routinely evaluated for all births. There was good agreement¹⁶ between reviewers (Kappa=0.87, 95% CI 0.74–1.00).

Second, the same two blinded reviewers performed a chart review of the EHR, examining the clinical notes and hospital-based autopsy reports to determine whether there was discordance between the clinical and pathologic diagnoses. In this jurisdiction, Medical Examiners’ reports are not public records and were not available for evaluation. The Goldman classification for clinical-pathologic discordance was used,¹⁷ which has been cited extensively in the autopsy literature.^18–20 This classification system assigns a number based on the extent of clinical-pathologic discordance. Classes 1 and 2 represent “major discordance”, Classes 3 and 4 represent “minor discordance,” and Class 5 represents no discordance. Table 1 illustrates examples of each class. An example of a claim with major discordance was a patient diagnosed with sepsis at time of death but on autopsy was found to have a 1.5cm valvular vegetation and would have qualified for surgery had it been discovered on echocardiography prior to death. An example of a claim with minor discordance is a patient who died in the setting of a large stroke who on autopsy was also noted to have small infarcts in the liver and spleen. There was very good agreement between reviewers (Kappa=0.97, 95% CI 0.94–0.99). When the first two reviewers did not agree, adjudication for the Goldman classification was performed by a third reviewer who is a board certified pathologist (JBI). Two patients’ EHRs did not contain clinical information because the events occurred before clinical notes were incorporated into the EHR, and thus could not be evaluated.

Table 1:

Goldman classification for clinical-pathologic discordance with examples

Goldman class	Clinical-Pathologic Discordance	Definition	Example
1	Major	Major diagnosis at autopsy that had not been made clinically AND detection before death would in all probability have led to a change in management that might have resulted in cure or prolonged survival	Clinical diagnosis: sepsis Autopsy diagnosis: newly diagnosed 1.5cm valvular vegetation who would have qualified for surgical intervention for endocarditis has it been found on echo
2	Major	Major diagnosis at autopsy that had not been made clinically BUT detection before death would probably not have led to a change in management	Clinical diagnosis: pneumonia and acute respiratory distress syndrome (ARDS) Autopsy diagnosis: newly diagnosed pulmonary emboli however anticoagulation would unlikely have altered patient’s course given patient’s underlying lung cancer and ARDS
3	Minor	Missed minor diagnoses that were diseases related to the terminal disease process but were not directly related to death	Clinical diagnosis: large stroke Autopsy diagnosis: small infarcts in liver and spleen
4	Minor	Missed minor diagnoses that were either important unrelated diagnoses that might eventually have affected prognosis or processes that contributed to death in a terminally ill patient	Clinical diagnosis: myocardial infarction requiring cardiopulmonary bypass in patient with rising lactate prior to death Autopsy diagnosis: bowel ischemia
5	None	Clinical-pathologic agreement	Clinical diagnosis: hemorrhage from a ruptured abdominal aortic aneurysm Autopsy diagnosis: hemorrhage from a ruptured abdominal aortic aneurysm

Exposure

The primary exposure was a binary variable of autopsy being performed regardless of whether it was performed at the hospital or by the Medical Examiner. For the subgroup of claims in which the autopsy was performed at the hospital, the exposure was the binary variable of major versus minor/no discordance between clinical and pathologic findings.

Outcome

The primary outcome was percent of paid claims and the amount paid. A claim was considered paid if the compensation paid on behalf of any defendant was non-zero. The claim could be paid via settlement or plaintiff verdict.

Statistical Analysis

Patient-, provider- and claim-related characteristics were compared on the basis of 1) whether an autopsy had been performed and 2) whether there was major versus minor or no clinical-pathologic discordance for cases in which an autopsy was performed. Definitions of these characteristics are described in the eMethods. Wilcoxon rank sum tests were used to compare medians and Fisher exact tests were used to compare percentages. Indemnity amounts were adjusted to December 2017 valuation using the Consumer Price Index taking into account the month and year in which the claim closed.²¹

SAS 9.4 (SAS Institute, Cary, NC) was used for statistical analysis. The threshold for significance was P = 0.05.

Subgroup analysis

We compared select characteristics of claims in which the autopsy was performed by the hospital pathology department versus the Medical Examiner because claims involving a Medical Examiner may more likely be due to errors on behalf of the hospital or provider, which may increase the likelihood of these claims resulting in payment.

Post hoc sensitivity analysis

We performed median regression analysis to corroborate the significant association found on univariate analysis between having an autopsy performed and compensation amount. We adjusted for the number of defendants as an ordinal variable because we found in earlier analyses that there were more defendants in claims with autopsies, which could increase the total compensation paid.

Results

Of all claims involving an inpatient death with loss date between 1995–2015 (n=293 claims), autopsies were performed in 89 (30%) claims, either by the hospital (69% of autopsies) or Medical Examiner (31% of autopsies) (Figure 1). Of claims in which a hospital-based autopsy was performed, 18 (31%) had major clinical-pathologic discordance.

Figure 1:

Flow chart of claims used in the study

Table 2 shows the patient-, provider- and claim-related characteristics by whether an autopsy was performed. A major surgical procedure was involved in 46% (n=135) of inpatient death-related malpractice claims. Claims in which autopsies were performed had lower patient age at death (48.1 years versus 59.7 years, P<0.001). There was no difference in the percent of claims in which an autopsy was performed over the time period by five-year increments, which was confirmed by a Mantel-Haenszel trend test (P=0.93). The most common primary responsible providers for claims in which autopsies were performed were internal medicine (11%), obstetrics (11%), and general surgery (8%). The most common allegation category for inpatient death malpractice claims was “diagnosis-related”, which was less common in claims in which autopsies were performed versus not (18% versus 38%, P=0.001). Physician trainees were more likely to be named as defendants in claims in which autopsies were performed versus not performed (48% versus 33%, P=0.02). There was no difference in specific claim disposition (i.e. percent dropped/denied/dismissed versus settled versus defense verdict versus plaintiff verdict) depending on whether autopsies were performed or not. Additionally, there was no difference in the percent of paid claims depending on whether or not an autopsy was performed (42% versus 41%, P=0.90).

Table 2:

Characteristics of malpractice claims from inpatient death events at three academic centers by presence of autopsy data

	All claims resulting from inpatient deaths n=293	Claims with autopsy n=89	Claims without autopsy n=204	P value
Patient-related
Median patient age (IQR)	56.7 (38.1, 71.0)	48.1 (31.2, 63.5)	59.7 (43.5, 73.7)	0.0003
Patient male	141 (48%)	45 (51%)	96 (47%)	0.61
Loss year
1995–1999	72 (25%)	20 (22%)	52 (18%)	0.66
2000–2004	98 (33%)	31 (34%)	67 (33%)	0.79
2005–2009	86 (29%)	27 (30%)	59 (29%)	0.89
2010–2015	37 (13%)	11 (12%)	26 (13%)	1
Provider-related
Primary responsible provider
Internal medicine	32 (11%)	10 (11%)	22 (11%)	1
Obstetrics	29 (10%)	10 (11%)	19 (9%)	0.67
General surgery	25 (9%)	7 (8%)	18 (9%)	1
Claim-related
Major surgical procedure involved	135 (46%)	47 (53%)	88 (43%)	0.16
Final diagnosis from malpractice claim
Laceration during surgery	21 (7%)	8 (9%)	13 (6%)	0.46
Lung cancer	10 (3%)	0	10 (5%)	0.04
Issue with anticoagulants	6 (2%)	3 (3%)	3 (1%)	0.37
Allegation
Diagnosis-related	93 (32%)	16 (18%)	77 (38%)	0.001
Improper performance of surgery	26 (9%)	9 (10%)	17 (8%)	0.66
Improper management of surgical patient	25 (9%)	12 (13%)	13 (6%)	0.07
Median filing time, years (IQR)	2.1 (1.1, 3.0)	1.9 (1.1, 3.0)	2.2 (1.1, 3.0)	0.08
Median open claim time, years (IQR)	3.7 (1.9, 4.8)	3.8 (2.1, 4.7)	3.7 (1.9, 4.9)	0.86
Median number of defendants (IQR)	1 (1, 3)	2 (1, 3)	1 (1, 2)	0.03
Staff physician named as defendant	232 (79%)	72 (81%)	160 (78%)	0.75
Physician trainee named as defendant	111 (38%)	43 (48%)	68 (33%)	0.02
Contributing factors
Patient assessment issues	174 (59%)	49 (55%)	125 (61%)	0.37
Selection and management of therapy	98 (33%)	33 (37%)	65 (32%)	0.42
Communication among providers	85 (29%)	25 (28%)	60 (29%)	0.89
Communication between patient/family and providers	71 (24%)	21 (24%)	50 (25%)	0.12
Patient monitoring	65 (22%)	29 (33%)	36 (18%)	0.006
Disposition
Dropped/denied/dismissed	122 (42%)	33 (37%)	89 (44%)	0.31
Settled	111 (38%)	33 (37%)	78 (38%)	0.90
Defense verdict	56 (19%)	21 (24%)	35 (17%)	0.20
Plaintiff verdict	4 (1%)	2 (2%)	2 (1%)	0.59
Claims paid	120 (41%)	37 (42%)	83 (41%)	0.90

n (%) is reported except if otherwise specified. P-values were obtained comparing claims with and without autopsy using Fisher Exact test for comparisons of a percent and Wilcoxon rank sum test for comparison of a median. Zero organizations were named as defendants. IQR=interquartile range.

Univariate and median regression analyses were performed to evaluate amount of compensation paid depending on whether an autopsy was performed. On univariate analysis, the median indemnity was higher in claims in which autopsies were performed ($1,180,537 versus $906,518, P=0.02). After adjusting for the number of defendants, however, there was no longer a statistically significant association (effect estimate $326,542, 95% CI -$116,451 to $769,535, P=0.11).

Supplemental Table 1 shows select characteristics of claims in which autopsies were performed by a hospital’s pathology department versus an outside examiner. Of the following claim-related factors (open claim time, most common final diagnosis, most common provider involved, most common allegation, claim disposition, claim payment and indemnity amount for paid claims), the only difference was that claims in which autopsies were performed by hospitals were more likely to have a diagnosis-related allegation (20% versus 15%, P=0.005).

Table 3 demonstrates characteristics of the 59 evaluable claims in which hospitals autopsies were performed, split by whether there was major or minor/no clinical-pathologic discordance. There were no statistically significant differences in final diagnosis, allegation category, or claim disposition between the two groups. In addition, there was no increase in percent of paid claims (44% versus 41%, P>0.99) or median indemnity ($895,954, IQR $622,504-$1,571,389 versus $1,494,120, IQR $1,036,521-$2,371,369; P=0.10) when there was major discordance between the clinical and pathologic cause of death, as compared to minor/no discordance.

Table 3:

Characteristics of malpractice claims in which an autopsy was performed based on the Goldman Classification of clinical/pathologic discordance

	Claims with major discordance n=18	Claims with minor or no discordance n=41	P value
Median claimant age (IQR)	52.3 (37.3,66.9)	43.7 (23.8, 60.7)	0.26
Claimant male	9 (50%)	24 (58%)	0.58
Primary responsible provider
General surgery	3 (17%)	4 (10%)	0.66
Obstetrics	<2	6 (15%)	0.42
Major surgical procedure involved	8 (44%)	28 (68%)	0.15
Median open claim time (IQR)	3.3 (1.9, 4.1)	3.7 (2.1, 4.7)	0.33
Final diagnosis from malpractice claim
Laceration during surgery	<2	6 (15%)	0.16
Central nervous system complication	2 (11%)	<2	0.22
Iatrogenic pulmonary embolus	<2	2 (5%)	>0.99
Allegation
Diagnosis related	3 (17%)	8 (20%)	>0.99
Improper performance of surgery	2 (11%)	6 (15%)	>0.99
Improper management of surgical patient	3 (17%)	4 (10%)	0.66
Disposition
Dropped/denied/dismissed	6 (33%)	14 (34%)	>0.99
Settled	7 (39%)	16 (39%)	>0.99
Defense verdict	5 (28%)	10 (24%)	0.76
Plaintiff verdict	<2	<2	>0.99
Paid	8 (44%)	17 (41%)	>0.99
Median indemnity payment, dollars (IQR)	$895,954 ($622,504, $1,571,389)	$1,494,120 ($1,036,521, $2,371,369)	0.10

Claims are divided into columns by whether there was deemed to be major or minor/no clinical-pathologic discordance regarding the cause of and diagnoses at death as assessed by the Goldman classification for clinical-pathologic discordance. n (%) is reported except if otherwise specified. P-values were obtained using Fisher Exact test for comparisons. A cell that contains <2 could be 0 or 1 and is marked as such to ensure claims are not identifiable.

The claim types, allegations, and diagnoses for claims in which there was major clinical-pathological discordance are illustrated in Figures 2A–C. The majority of claims were medical (56%) versus surgical (44%). The most common allegation was improper medical treatment (33%). The most common diagnoses discovered on autopsy were infection or sepsis (20%), pulmonary or air embolus (20%) and coronary atherosclerosis (12%).

Figure 2A-C:

Pie charts of A) type B) allegation and C) diagnosis of malpractice claims in which there was major clinical-pathologic discordance according to the Goldman Classification

A) The sample size of claims with major clinical/pathologic discordance was 18.

B) One allegation was assigned per claim.

C) More than one diagnosis could be assigned to a claim if there were more than one diagnosis discovered on autopsy that was not known clinically at death.

Discussion

We performed a multicenter study of medical malpractice claims from a captive insurer using a novel linkage to the EHR and found no association between autopsies being performed and the likelihood of having a paid malpractice claim. After adjusting for the number of defendants, we similarly did not find an association between compensation amount and autopsy being performed for paid claims. Lastly, we found that a sizable fraction (31%) of claims with hospital-based autopsies had major discordance between clinical and pathologic findings at the time of death. For claims with major clinical-pathologic discordance, there was no difference in either percent of paid claims or amount of compensation paid. We believe this is the largest study published to date examining the association between autopsies and medical malpractice outcome.

Our results extend previous studies, which have predominantly examined the pediatric population. A study of 99 appellate court cases between 1970–2002 demonstrated that most claims in which autopsies were performed contained the autopsies results in the claim.¹¹ A study in the same series showed that diagnostic error was uncommon in claims in which physicians were found to be negligent.¹⁰ They describe that treatment falling below the standard of care was more often associated with plaintiff verdict than new diagnoses discovered on autopsy.¹⁰ Our data are more recent, contain claims from patients of any age whether or not they were litigated, and link to the EHR to access the original autopsy reports. Notably, we found that claims without autopsies were more likely than claims with autopsies to have a diagnostic-related allegation, demonstrating a potential important role for an autopsy to provide diagnostic clarity for families in these cases.

We elected to include autopsies performed by a Medical Examiner or outside examiner in the autopsy group but recognize that these patients could be different compared to patients receiving autopsies at the hospital. For example, in Massachusetts, Medical Examiners request autopsies of any patient who died from alcohol or drugs or was <18 years of age. In addition, a Medical Examiner may request an autopsy if there is concern regarding the medical care provided, such as death in the emergency department or within 24 hours of hospital admission. Autopsies in these scenarios may uncover a missed, acute diagnosis that may be the result of an error, which could mean that Medical Examiner cases are more likely to result in a paid malpractice claim outcome. We thus performed sensitivity analysis comparing select characteristics of claims in which the autopsy was performed by the hospital versus the Medical Examiner. We confirmed the groups were overall fairly similar. The one statistically significant difference of diagnosis-related allegations being more common in claims in which autopsies were performed by hospitals may be due to lack of power.

We studied an extended period of time to maximize the number of claims in the cohort. However, there were several trade-offs as a result. First, claims were included regardless of the medical malpractice milieu in Massachusetts. The medical malpractice system underwent reform in 2012 with changes to the disclosure/apology law, charitable caps on institutions and a notice period of intent to file a claim.²² Second, the rate of autopsies being performed decreased over the time period.²³ Within our cohort, however, there was no difference in the proportion of claims with autopsies over time by five-year increments. Although this quantitative trend over time was somewhat reassuring, the reasons for obtaining an autopsy could have changed over time. For instance, patients in the early part of the cohort could have undergone autopsies to clarify diagnoses that remained unclear prior to routine use of advanced imaging and imaging-guided diagnostic procedures. In addition, patients in the latter part of the cohort could have undergone autopsies so that families and plaintiff attorneys could document missed diagnoses to utilize in their claim. Our analyses are limited in their ability to evaluate the decision-making process around pursuing autopsy.

Likewise, the results of this study must be interpreted in the context of the study design. First, given this study involves a rare event, we needed to pool several hospitals’ and years of data. Despite doing this and being able to link most claims, the sample size of malpractice claims involving inpatient deaths was only 293, which narrowed to only 59 hospital-based autopsies. This restricted us from performing more detailed multivariable regression analyses on the subgroup of patients with hospital-based autopsies. Second, based on information available, we focused this study on claim outcome, not filing of claims. This study does not address whether inpatient deaths with autopsy were more likely to have a claim filed in the first place. This is an area of further study that our group plans to pursue, and involves obtaining clinical records for thousands of inpatient death cases and syncing these data with all claims filed at the respective institutions. A third limitation is that all three hospitals were within the same state, which means that the results are generalizable to Massachusetts but may not be generalizable in states with different medicolegal milieux. Fourth, not all autopsies performed are exhaustive and families can decide which organs are or are not examined, potentially skewing the pathologic findings. However, based on our chart review, almost all of the autopsies examined most tissues in the body. Fifth, autopsies may be pursued more often when there is more diagnostic uncertainty. However, this does not necessarily reflect that the families’ intent at the time of autopsy consent was to use it as part of the litigation strategy. In addition, pathologists are informed by clinical notes and radiographic findings, so may be inherently predisposed to look for certain conditions on autopsy that are suggested by the clinical understanding. Therefore, even though autopsies represent the diagnostic gold standard, diagnoses may be missed if pathologists are misguided by clinical suspicion. Finally, we noted that there was a slightly higher indemnity payment when an autopsy was performed which was related to the number of defendants involved. Although there was no difference in median indemnity payment after adjusting for the number of defendants, the finding that more defendants were involved in cases in which an autopsy was performed remains unexplained.

As we aspire to be a learning health system,²⁴ we must continuously leverage data on past performance to treat future patients in a more informed way. Autopsy results are an important source of data for many stakeholders, including clinicians, trainees, patients, and families. Despite its limitations, this study provides evidence that performance of an autopsy is not associated with an increased risk of claim with payment or inflation-adjusted compensation of paid claims, even when there is major discordance between the autopsy and clinical diagnoses at the time of death. Liability fears should not be seen as a barrier to obtaining autopsies. We recommend that the decision to ask families for an autopsy be based on the potential value of learning for the family and the clinicians involved in the patient’s care.