Cause-of-death reporting enables health departments determine program needs and effectiveness. Because clinicians often report cause of death incompletely or inaccurately, cardiovascular disease and pneumonia have been over-reported in some regions (1–5). The National Center for Health Statistics (NCHS) recommends that health departments query clinicians after death registration to obtain more detailed and accurate information (6,7). New York City’s Department of Health and Mental Hygiene (DOHMH) pilot-tested an electronic system for querying clinicians about cause of death and asking for more information when unspecified pneumonia, which is usually precipitated by a chronic condition, was reported as the underlying cause.

We selected a 519-bed acute tertiary-care facility that reported substantially more unspecified pneumonia deaths than other New York City hospitals. In 2010, 16% of deaths at this hospital were reported as due to pneumonia or influenza, compared with 6.8% citywide.

In New York City, 93% of 2012 deaths were entirely reported through the Electronic Vital Events Registration System (EVERS) (8). In EVERS, the text provided by certifying clinicians on each death certificate is used to determine the underlying cause of death and to assign the corresponding International Classification of Disease (ICD-10) code (9). A standardized international algorithm is used to apply these ICD-10 codes either automatically by NCHS-provided software or manually by our nosologist if automated coding fails or is unavailable (10).

Beginning March 1, 2012, we queried all deaths registered from January 1 through July 31, 2012, that were coded J189 (pneumonia, unspecified) as the underlying cause of death. Our DOHMH nosologist sent an email and system message via EVERS to certifying clinicians and copied the hospital administrative staff member responsible for death registration (Box). The message asked clinicians to submit electronically, through EVERS, either an amendment with a revised underlying cause of death or a comment within the death record stating that additional information was unavailable.

The New York City health code mandates responses to DOHMH information requests within 5 business days (11). Nonresponses were followed with an email and an EVERS message after 2 weeks and with a telephone call after 4 weeks.

DOHMH quality assurance staff reviewed clinician amendments and comments. Our nosologists applied ICD-10 coding to approved amendments by using the same methods used during death registration. For rejected amendments, quality assurance staff or DOHMH physicians called the clinician to discuss the case.

A DOHMH physician trained in cause-of-death documentation who was temporarily assigned to the Bureau of Vital Statistics reviewed medical records of queried cases to determine the correct cause of death. The nosologist then manually assigned codes to the corrected underlying causes for comparison with original and amended causes reported.

Of the 29 death certificates we queried (4.6% of the 633 death certificates registered by the hospital from January 1 through July 31, 2012), 12 (41.3%) required a second message because of nonresponse. Clinicians responded to 22 queries (75.9% of all queries) by August 31, 2012, which was the end of the 4-week follow-up period for the last query sent. After further outreach to administrators, another 6 query responses were received by October 16, 2012, for a final response rate of 96.6%. All amendments were reviewed and accepted electronically.

Of the 28 amended cases, 27 (96.4%) amendments led to a change in the underlying ICD-10 code to one other than J189 (Table). Twelve (42.9%) amendments were inadequate in that they provided only more specific information about the infecting organism rather than providing information about the medical condition(s) that made the patient vulnerable to acquiring and dying of pneumonia.

The DOHMH physician review found that the medical records related to all 29 queried deaths contained information about the cause of death that would have changed the underlying cause reported (Table). Underlying causes based on DOHMH review matched the underlying cause-of-death codes based on the clinician amendment in only 3 cases (10.7%). For 2 cases, no evidence was found that pneumonia was involved in the death at all. Additionally, the reviewing physician found that in 2 instances the certifying clinician should have reviewed the case with the medical examiner’s office because injury contributed to the death.

To our knowledge, this is the first study to document feasibility of querying cause of death, an NCHS-recommended best practice (6), using an electronic death registration system. Most health departments use time-consuming postal mail to query, requiring manual data entry of updated causes of death. In our study, electronic querying was operationally feasible, minimizing time between death and query initiation, reaching clinicians when they remembered case details better and while they were still at the reporting hospital, a key challenge in New York City where many are physicians-in-training who rotate to other locations. Nonetheless, query compliance and accuracy were major challenges.

Although the electronic messages emphasized that New York City’s health code required a prompt response, certifying clinicians had a low initial response rate. On follow-up, we found that nonclinical hospital staff members were unable to compel clinicians to respond. Therefore, we met with hospital medical and administration personnel, including medical department heads, to encourage their support. The administration expressed support for the querying system, agreeing to designate a clinician “champion” to facilitate timely query responses. Following this meeting, more queries were returned, but response quality, based on medical chart audit, was low. Amendments did not adequately address the underlying medical condition that put patients at risk for contracting or dying of pneumonia and were often inaccurate. Some inaccuracy probably occurred because the facility assigned a chief resident to respond to all queries instead of having the certifying clinician respond about his or her own certificates.

Our findings highlight the continued need to enforce New York City requirements for accurate cause-of-death information. We have disseminated new clinician education materials in the pilot hospital, updated a citywide online training module (12,13), and converted a public health epidemiologist’s position at DOHMH to a part-time physician position. The physician will conduct in-service trainings, follow up on unreturned queries, and audit medical charts to enable issuance of formal citations for inaccurate cause-of-death reporting. Given the limited time allocated to the task by a part-time position, the physician will focus on hospitals with large discrepancies between their billing discharge information and submitted cause of death based on assigned ICD-10 codes. We plan to expand this electronic death query system on a hospital-by-hospital basis with modifications to address noted challenges and will include a meeting with hospital administration prior to implementing the intervention to involve them in the effort.