Death Certification Errors at an Academic Institution

The death certificate is a legal document with diverse and far-reaching applications. It is the final certification of a patient's life and cause of death (disease or injury), provides invaluable information to the decedent's family, and is required for most end-of-life legal matters such as life insurance and inheritance. It also contains epidemiologic data that are essential for formulating vital statistics and allocating public health resources.

Given the importance of this document, it is crucial that its completion be thorough and accurate. Unfortunately, errors in death certification are common1–10 and range from incomplete certificates and illegible handwriting to inaccurate causes and manners of death.3,7,9 Also seen is the use of medical abbreviations7 rendering the certificate unintelligible to the general public.

Of the relevant published studies, some focus solely on specific disease entities,2,4,6 and only a few separate errors into categories such as major and minor errors.3,7,9 Categorization is crucial for recognizing the types of errors commonly made so that methods may be designed to prevent them. The objective of this study was to determine the frequency and types of death certification errors made by clinicians at an academic medical center, based on a review of the information available in the medical charts.

Fifty random patients were selected who died at this institution between January 2002 and December 2003 and did not undergo an autopsy. A medical chart review was undertaken for each case, and a brief clinical summary (approximately 50–100 words) was produced by one of us (B.S.P.) that consisted of information from the physician notes and relevant laboratory and radiology results. This clinical summary was an objective condensation of the patient's medical chart and did not include a clinical impression of why the patient died. Two pathologists with extensive experience in death certification (N.J.H. and S.L.S.) used this summary to independently create mock death certificates while blinded to the content of the original certificate. Each pathologist then compared the original with the mock death certificate and identified the number and type of errors in the original certificate.

Errors were assigned a grade on a I to IV scale of increasing severity (Table 1). A grade of zero was assigned if no error was identified. Error grades I and II included minor errors such as incompleteness, illegibility, illogical or nonsensical order, and minor missed comorbidities. Error grade III included major missed comorbidities or major missed immediate or intermediate causes of death. Error grade IV, the most severe, included omitted or wrong cause of death or wrong manner of death. These latter error grades (III and IV) are mistakes or omissions that could significantly change the death certificate interpretation by the family, other physicians, or personnel gathering heath statistics. Cases with multiple errors received more than 1 grade when appropriate.

Discrepancies in grade and number of errors were resolved by consensus review involving both pathologists and a third party (B.S.P.). This third party did not participate in the generation of the mock death certificates but assisted in judgments concerning the grade (severity) of error.

During the 2-year study period, there were 1301 deaths that occurred in the medical center. Of these, 955 patients (73%) did not undergo institutional or medical examiner autopsy and therefore had death certificates completed by an attending clinician or house officer. In this institution, a pathologist completes the death certificates for patients undergoing autopsy.

Following review and comparison of the pathologist-completed mock death certificates, consensus was reached in 100% of the cases. Differences between the 2 grading pathologists primarily occurred not in the mock death certificates but in the error grades assigned to each case. Grading differences consisted of overlooked errors on the part of one or the other pathologist and were easily identified during the evaluation and discrepancy review process. There were no ultimate disagreements among the 2 pathologists or the independent third party.

Of the 50 clinician-certified death certificates reviewed for this study, 48 (96%) contained some degree of error (Table 2). The percentages of grade I, II, and III errors were 72%, 32%, and 30%, respectively. The most common error was grade Ia (52%). Seventeen cases (34%) received the highest (grade IV) error score, with 13 (26%) IVa and 4 (8%) IVb errors. Multiple errors were common, with 82% containing errors from more than 1 category.

Of the 50 original certificates, 43 (86%) were marked as resident certified and 7 (14%) as attending certified by a checked box on the certificate. Because of illegible signatures and handwriting, the identity of the resident and his or her postgraduate year could not be ascertained in most cases. Of the 7 attending-certified cases, 2 had a grade IV error (both IVa), 3 had only a grade I or II error, 1 had a grade III error, and 1 had no errors.

When completing a death certificate in the state of Vermont, the physician is responsible for several categories, including the name of the patient, the date and time of death, the cause and manner of death, and any contributing factors. The physician must also specify whether tobacco was suspected to have contributed to the patient's death, whether an autopsy was performed, and whether autopsy results were available at the time of certification. Finally, the physician must sign the form and print his or her name legibly. Instructions for writing the cause-of-death statement are readily available,11,12 and some are printed on the certificate. Despite the limited amount of data required from a physician, there are many opportunities for error, as shown by the results of our study.

Death certification errors are a universal problem, and reported major error rates at other institutions range from 24% to 37%.3,7,9,10 It is difficult to directly compare these results with those of our study because of variation in the criteria used to define a major error. However, there is uniform agreement among most of these studies, including ours, that the wrong cause or manner of death and a lack of an acceptable underlying cause of death qualify as major errors. These are errors that, if discovered, would necessitate rewriting or reissuing the death certificate. In our study, 34% of the 50 certificates contained such major errors (wrong cause or manner of death) and hence were classified as grade IV. We subdivided grade IV errors into those that listed no underlying cause of death (grade IVa) and those that listed the wrong underlying cause or manner of death (grade IVb). Of the 2, grade IVa errors were more common and were seen in 26% of our cases. An example of a grade IVa error identified in our study cases is seen in Figure 1.

Of our 4 cases with grade IVb errors, 3 listed the wrong cause of death (based on the information gained from the medical chart review), and 1 listed the wrong manner of death. An example of a wrong cause of death from our study is the case of a 74-year-old man with pulmonary fibrosis due to chemotherapy for thyroid cancer and ulcerated Barrett esophagus. Before death, the patient presented with large-volume esophageal hemorrhage and hypotension. Despite administration of packed red blood cells and vasopressors, the patient became increasingly hypotensive and developed respiratory insufficiency. Because of a preexisting do-not-intubate order, he was placed on comfort care and died the following day. The intern on duty certified the cause of death as pulmonary fibrosis due to chemotherapy due to thyroid cancer. However, both pathologists in this study thought that esophageal hemorrhage due to ulcerated Barrett esophagus due to gastroesophageal reflux disease was the correct cause of death sequence and that pulmonary fibrosis was a contributing factor. This demonstrates the difficulty in deciding the cause of death when multiple possible causes exist. Ultimately, the certifying physician must use best medical judgment to make this determination.

The single case of a grade IVb error (wrong manner of death) was an older patient with breast cancer who fell and sustained a subarachnoid hemorrhage. She died following a short course in the intensive care unit, and her manner of death was listed as natural. However, her known breast cancer did not adequately account for her fall. Therefore, this case should have been properly classified as an accidental death. This is an important distinction because accidental deaths usually require investigation by the office of the chief medical examiner or coroner.

Determining and designating some of the lesser grade errors (grades I-III) also involved a certain degree of subjectivity. For example, exclusion of comorbidities was classified as minor (grade II) or major (grade III) based on clinical history; hypertension was considered a minor comorbidity in a patient who died of metastatic colon cancer but a major comorbidity in a patient who died of an acute myocardial infarct. This distinction could not always be made with certainty, and, in cases of doubt, a conservative classification (grade II error) was chosen.

In addition to major missed comorbidities, the grade III error category included major missed components of the cause-of-death sequence (immediate or intermediate causes); in other words, an important part of the “story” was missing. For example, listing adenocarcinoma of the colon as the cause of death without mentioning that the patient had died following partial colectomy, complicated by recurrent pulmonary emboli and lung infarction, is omitting a major portion of the patient's course.

As well as minor missed comorbidities, a grade II error was assigned when a diagnosis was not as specific as possible. For example, a diagnosis of lung cancer was considered an error if a more specific diagnosis of non–small cell carcinoma or squamous cell carcinoma of the lung was available. This type of nonspecificity was common in our study, and it is debatable whether it should be considered an error. However, this degree of specificity is essential for epidemiologic studies and may be useful for family records. Furthermore, we are of the opinion that a physician should be familiar with his or her patient's specific diagnoses and transfer them to the death certificate.

Grade I errors included missing physician contact information (grade Ia) and illegible signatures (grade Ib). Both made it difficult, if not impossible, to track down the certifying physician, should a question or problem arise. Abbreviations were also common (eg, CABG, HTN, DM II, AMI, and OA), despite the fact that the death certificate clearly states that they should not be used. Although this institution has a list of approved abbreviations for internal use, these abbreviations are not necessarily used by other health care institutions and are not generally recognizable by the layperson or family member. A grade Ic error was assigned when part I diagnoses were not listed in a logical order on the death certificate. For example, listing coronary artery disease on the first line, followed by acute myocardial infarct on the second line, is incorrectly stating that the coronary artery disease was due to the infarct. In general, grade I to grade III errors did not necessitate rewriting or reissuing the death certificate.

There are many probable reasons for our high (96%) overall error rate. First and foremost, our criteria were strict and rigidly upheld during error grade assignment. We checked for the use of abbreviations, illegibility, failure to check all appropriate boxes, and writing in the wrong place (all grade I errors). Although these errors may not significantly misrepresent the underlying cause of death, they signify a degree of carelessness, haste, or lack of experience with death certificates on the part of the physician (Figure 2). Therefore, we believe it is significant that 72% of all certificates in our study contained a grade I error.

Based on our experience, additional reasons for error include physician inexperience, fatigue, time constraints, unfamiliarity with the medical history of the deceased, and perceived lack of importance of the death certificate. In this institution, the task of death certification usually falls to the least experienced member of the physician team, the intern. This individual may not fully understand the importance of the death certificate. In many cases, the only death certification training a house officer receives in this institution is a 15-minute presentation at the time of orientation. Furthermore, constraints on the number of residents and the permitted work hours make it common for one on-call resident to cover up to 80 patients during a night shift. A busy intern who is cross-covering services may lack the motivation or the time to review the medical chart of an unfamiliar patient to accurately formulate the cause-of-death sequence. It is expected that a house officer will gain experience in death certification throughout his or her residency, but without proper training and reinforcement, bad habits may be acquired. Of note, there was no appreciable difference in the grade and number of errors made by attendings compared with residents; however, the number of attending-certified deaths (n = 7) in this study is too small to have statistical significance.

To counter the numerous factors that adversely affect proper death certification, we suggest a multifaceted approach. First, we recommend a required annual course in death certification for all house staff to introduce and reinforce proper techniques. Myers and Farquhar7 demonstrated a 15.7% decrease in the error rate following a 1-time educational intervention. If repeated on a regular basis, the error rate may decrease further. During such a course, the importance of the death certificate and examples of commonly encountered cases and errors can be presented. Furthermore, an opportunity is provided to inform clinicians of the resources available to them, should questions arise regarding death certification, including the on-call pathology resident and, in appropriate cases, the office of the chief medical examiner or coroner. In this institution, 1 to 2 hours of death certification training is given in medical school, as well as 15 minutes during house staff orientation. Plans to institute a more in-depth annual course are under way. Furthermore, death certification packets with instructions are provided at all nursing stations as an educational resource for house staff and attending physicians.

To address the problem of unfamiliarity with the clinical history of the deceased, we recommend that death certification be completed by a physician who is familiar with the patient, even if this delays its completion until the morning. In our experience, it is rarely necessary for a death certificate to be completed during the night shift. We also propose that the death certificate be discussed the following day during rounds with the entire physician team. Ideally, the attending would be involved in this discussion, because Vermont state law maintains that the attending physician is ultimately responsible for completion of the death certificate. A pathologist or pathology resident could also participate in this discussion. This would allow ongoing education of the house staff and provide an opportunity to revisit the patient's hospital course. If, after careful consideration, the cause of death cannot be determined with certainty, or if multiple possible causes of death exist, we encourage the clinicians to use their clinical judgment and determine the most likely cause of death based on the patient's history and presentation.

To strengthen and augment their death certification skills, it has been suggested that physicians need continuous feedback.5 The practice in this institution (from 2004 to the present) is that every death certificate signed in the state of Vermont is reviewed by the office of the chief medical examiner. When egregious (grade IV) errors are identified, the certifier is notified by telephone and is asked to recertify the death or agree to have the death certified by the medical examiner.

In summary, it is evident that the error rate of death certification in this and other reported studies1–10 is high and that this problem is faced by many institutions. When considering this error rate, it is crucial to recognize that this study includes only death certificates that are completed based on condensed clinical information without the benefit of autopsy. We consider these to be important data, given the low autopsy rates in the United States and elsewhere, but the superiority of the autopsy for determining the most accurate cause and manner of death is well recognized13–18. Although pathologists are not infallible, it is our opinion that, by working with clinicians, pathologists are best suited to assess the patient, perform a postmortem examination, and complete the death certificate. Of note, the 2 senior pathologists brought diverse education and experiences (hospital versus forensic autopsy training, for example) to this study and still had almost complete concordance in completion of their mock death certificates. Although colleagues often adopt similar practices over time, these pathologists had almost no experience working together before this study. In lieu of a pathologist-certified death, the pathologist may aid the certifying clinician by providing education and consultation.