Discussion
This is the first comprehensive survey of medical, surgical, and emergency medicine residents on the current cause-of-death reporting system. It is also the first to document that physicians completing death certificates often knowingly complete them inaccurately.
Most physicians responding to our survey perceived death certificates to be inaccurate. Physicians who completed 11 or more death certificates in the last 3 years (ie, physicians more experienced in cause-of-death reporting) had a more negative opinion of the accuracy of death certificates than physicians who completed 10 or fewer death certificates (ie, less experienced physicians). Residents reported being instructed to report causes of death they did not agree with by personnel (eg, hospital admitting personnel, medical examiner) not directly involved in the patient's care.
We identified several areas in the current system that may lead to inaccurate cause-of-death reporting. Consistent with results of prior studies, we found that residents overreported cardiac disease as a cause of death. Other causes of death, such as septic shock and acute respiratory distress syndrome, may be underrepresented. The cause-of-death reporting system in New York City does not recognize all symptoms or diagnoses as a cause of death. For example, a clinician cannot report septic shock as a cause of death unless the cause of septic shock (eg, Escherichia coli urinary tract infection) is identified. Our survey respondents recognized the limited number of diagnoses accepted by the system.
Previous studies explored interobserver variability in identifying cause of death on death certificates and attributed potential error to subjectivity of the medical practitioners. These studies concluded that better education was required to increase accuracy of death certificates. Although our study reinforces the idea that inaccuracy in cause-of-death reporting may arise from inadequate training of physicians, it also identifies systemic barriers that limit the ability of residents to report what they understand to be the most likely cause of death.
This study had several strengths. It was conducted in New York City, which is a large area and has many residency programs and hospitals that use the same cause-of-death reporting system. We were able to evaluate a large group of physicians completing death certificates within the same system. Because cause-of-death reporting systems are fairly similar throughout the United States, our findings may have broad relevance.
This study also had several limitations. Participation in our survey was voluntary, so several potential sources of bias exist. Although we sent the survey to every internal medicine, emergency medicine, and surgical residency program director in New York City, not all program directors may have forwarded the survey to their residents, and not all residents chose to respond. We could not confirm that all program directors forwarded the survey to their residents, but we were able to identify each program that responded. Residents who felt particularly negative or positive about death certificates may have had more or less interest in participating. Additionally, our survey was subject to recall bias; residents may have recalled only strongly positive or negative experiences, not accounting for the many times the system worked smoothly.
We have identified potential sources for improvement in the current cause-of-death reporting system. We suggest expanding the acceptable causes of death to all inpatient diagnoses codes and improving the training of resident physicians. We also recommend that residency programs review the way patient care is transferred between residents to ensure that the most effective signout processes (transfer of care at the beginning and end of a resident's hospital shift) are used to limit death certificate errors.
Residents need better training in proper completion of death certificates, including cause-of-death identification, when and why causes should be amended, and the implications of cause-of-death data for their community. Historically, residents have not been well educated as to what they can and cannot put on death certificates, and most have not undergone formal training in death certificate completion. Although New York City has developed a mandatory online training module for physicians, only 21.5% of our survey respondents had completed the training module. We found no significant differences in responses between residents who completed training and those that did not. Forty percent of respondents reported receiving training through their residency program; respondents who received this training did not report the system as more accurate.
Only one-third of the physician residents in our study believed the current cause-of-death reporting system in New York City is accurate. Residents routinely reported diagnoses on death certificates that did not match their medical judgments. These errors may have lasting effects on the public health priorities of the community. Reform is needed both in the training and education of residents and in the system itself. We hope these findings will contribute to improvements in the cause-of-death reporting system and eventually more appropriate distribution of health care dollars.