Evaluating the Use of an Electronic Death Registration System for Mortality Surveillance During and After Hurricane Sandy: New York City, 2012

Abstract

Objectives. We evaluated the use of New York City’s (NYC’s) electronic death registration system (EDRS) to conduct mortality surveillance during and after Hurricane Sandy.

Methods. We used Centers for Disease Control and Prevention guidelines for surveillance system evaluation to gather evidence on usefulness, flexibility, stability, timeliness, and quality. We assessed system components, interviewed NYC Health Department staff, and analyzed 2010 to 2012 death records.

Results. Despite widespread disruptions, NYC’s EDRS was stable and collected timely mortality data that were adapted to provide storm surveillance with minimal additional resources. Direct-injury fatalities and trends in excess all-cause mortality were rapidly identified, providing useful information for response; however, the time and burden of establishing reports, adapting the system, and identifying indirect deaths limited surveillance.

Conclusions. The NYC Health Department successfully adapted its EDRS for near real-time disaster-related mortality surveillance. Retrospective assessment of deaths, advanced methods for case identification and analysis, standardized reports, and system enhancements will further improve surveillance. Local, state, and federal partners would benefit from partnering with vital records to develop EDRSs for surveillance and to promote ongoing evaluation.

New York City’s (NYC’s) Health Department introduced an electronic death registration system (EDRS) in 2005, allowing medical providers, funeral directors, and Health Department registration staff to voluntarily report deaths and access death records through a Web-based system. Since then, local mandates for EDRS use and timeliness resulted in nearly complete electronic reporting and accelerated the processing of death certificates and the availability of population mortality data.¹ Across the United States, 46 of 57 vital event jurisdictions have implemented EDRSs, creating new opportunities to use mortality data for the timely surveillance of emerging public health concerns.^2,3 For example, fatalities are a commonly used indicator of the public health impact of natural disasters, including extreme heat events, floods, tornadoes, and hurricanes.⁴ Recently, lead US health agencies have called for greater research and evaluation of disaster-related systems and processes to address challenges faced by government and public health agencies, hospitals and clinicians, and academic researchers.⁵ The evaluation of local systems after disasters is therefore critical for identifying and addressing knowledge gaps for future events.

On October 29, 2012, posttropical cyclone Sandy (Hurricane Sandy) made landfall approximately 100 miles south of NYC, causing a record-breaking storm surge throughout coastal areas of the city.⁶ Extensive flooding and wind damage caused widespread power outages, transportation shutdowns, medical facility and residential evacuations, and disruptions to health care access.⁷ Environmental risks persisted in the days and weeks following the storm, especially in residential buildings with damage to electrical, heating, elevator, and potable water systems. This raised concerns not only for direct-injury–related fatalities but also for all-cause deaths attributable to hazardous conditions. In response, the NYC Health Department used its EDRS to conduct mortality surveillance during and after Hurricane Sandy. Surveillance objectives were to (1) identify and characterize Sandy-related deaths, and (2) examine all-cause mortality during the storm and identify poststorm hazards. To assess performance of the disaster mortality surveillance system, we evaluated key system attributes, identified strengths and limitations, and made recommendations for system improvements.

METHODS

Figure 1 illustrates the NYC EDRS and disaster mortality surveillance system used during Hurricane Sandy. After pronouncing a death, a provider creates and medically certifies a record with the decedent’s name and sex and the date, time, place, and cause or causes of death. In the event of injury or sudden death, the provider must refer the case to the Office of Chief Medical Examiner (OCME), where causes and related circumstances are further investigated.¹⁰ On occasion, the OCME may certify the death with a pending cause of death, which is amended after investigations or postmortem studies.

FIGURE 1—

Flowchart of the electronic death registration and mortality surveillance system: New York City, 2012.

Note. ICD-10 = International Classification of Diseases, 10th Revision⁸; NYC = New York City; SQL = Structured Query Language.

^aReporting requirements are outlined in the New York City Health Code.⁹

^bThe preliminary record is composed of fields required at the time of medical certification, including the decedent’s name and gender and the date, time, place, and cause or causes of death.

Data entered by the medical provider or examiner comprise the preliminary death record, which is shown with the dotted line in Figure 1. After a record is started, a funeral director (or in indigent cases the medical facility or city mortuary) coordinates burial arrangements, enters the decedent’s demographic information, and submits the record to the Bureau of Vital Statistics (BVS). The BVS registration unit manually reviews the record to ensure completeness and proper investigation of injury-related deaths and registers the records in the EDRS. The NYC Health Code requires deaths to be medically certified within 24 hours and the complete record to be submitted for registration within 72 hours of death.¹⁰

The registered death record undergoes numerous steps before finalization. By the next business day, cause or causes of death are assigned International Statistical Classification of Diseases, 10th Revision (ICD-10)⁸ codes using the desktop version of the Mortality Medical Data System software, which is disseminated by the National Center for Health Statistics.¹¹ If the software fails to assign a code, BVS nosologists manually code the cause or causes of death. Decedent residential addresses are also geocoded (e.g., to identify census tract) and manually corrected if rejected. Paper records, accounting for less than 7% of deaths, are submitted directly to the BVS.¹ The registration unit immediately creates an electronic record with identifying information and the date and time of death. Additional fields from the paper record are entered within 2 days. Preliminary death records can be analyzed from the EDRS before registration with Specialized Structured Query Language.

Sandy Mortality Surveillance

Two days before the storm, as predictions of Sandy’s impact escalated, the BVS implemented previously established contingency plans. One registration unit employee remained posted in the Manhattan office during the storm; however, power outages that began that night forced the building to close for a week. In accordance with continuation of operation plans, the registration unit relocated the following morning to a Brooklyn satellite site, which routinely operates as a burial desk 2 days a week. Citywide transportation problems resulted in limited on-site staff. Other BVS staff reported to Health Department headquarters in Queens or local evacuation shelters. Directly following the storm, the OCME began e-mailing BVS information on Sandy-related fatalities. The chief medical examiner defined Sandy-related deaths as those that included injuries directly related to the actual environmental forces of the storm or the direct consequences of these forces (e.g., structural collapse).

To identify additional cases, BVS staff searched cause of death text fields for Sandy-related terms, which were selected from disaster mortality literature (data available as a supplement to the online version of this article at http://www.ajph.org).^12,13 In particular, staff used the text string search function PRXMATCH in SAS version 9.2 (SAS Institute, Cary, NC) to search cause of death fields, including part 1 (A–C), part II, and “How the Injury Occurred,” which is completed when the death has an external cause. Potential Sandy-related cases identified through this process, and from news reports and media inquiries, were referred to the OCME. The BVS also activated an EDRS pop-up notice to prompt referral of suspected Sandy-related deaths to the OCME. The notice remained active until January 4 and read as follows:

If Hurricane Sandy was the direct cause of this death or contributed to this death either directly or indirectly, please report the death to the NYC Office of Chief Medical Examiner (OCME) at ▪▪▪-▪▪▪-▪▪▪ before certifying the case. Please include cases involving cold stress or carbon monoxide exposure occurring in residences without heat. Do not continue with the case if OCME is taking ownership.

In addition to tracking Sandy-related deaths, the BVS created daily reports of deaths owing to all causes, which compared death counts from October 1 through December 31, 2012, with the average of 2010–2011 death counts. Excess mortality was determined on the basis of the percentage difference between deaths in 2012 and 2010 to 2011 averages. Additional summary reports stratified all deaths by age, cause of death, and flood proximity. These reports used a combination of preliminary death records and registered death records to maximize timeliness.

Evaluation Design

We used the Updated Guidelines for Surveillance System Evaluation from the Centers for Disease Control and Prevention to assess disaster mortality surveillance during and after Hurricane Sandy.¹⁴ To evaluate system performance, we gathered evidence on key attributes, namely usefulness, flexibility, stability, timeliness, and quality. We interviewed BVS staff to assess system usefulness and flexibility. We assessed stability by reviewing continuity of operations and calculating the proportion of records reported electronically versus on paper. We tested for significant differences in median certification and registration time between 2012 and 2010 and between 2012 and 2011 using the Wilcoxon signed-rank test. We also examined variation in timeliness by borough of report, Sandy-relatedness, medical examiner cases, and proximity of the decedent’s residential address to flooding (on the basis of the census tract from Federal Emergency Management Agency maps).¹⁵

To assess data quality, we compared the number of additional Sandy-related deaths identified by text string searches and media inquiries with deaths identified by the OCME. Finally, we compared the quality of selected variables collected on preliminary records with the final registered records. We calculated the percentage of records with complete information (i.e., no missing, blank, or unknown values) and the proportion of preregistration records that agreed with registered records by selected item.

RESULTS

Table 1 summarizes the key attribute definitions, strengths, and weaknesses of using NYC’s EDRS for disaster mortality surveillance during and after Hurricane Sandy.

TABLE 1—

Key Attributes, Strengths, and Weaknesses of Using New York City’s Electronic Death Registration System for Disaster–Mortality Surveillance: 2012

Evaluation Item	Definitiona	Strengths	Weaknesses
Usefulness	Contribution to the detection, prevention, and control of an adverse health-related event	Rapidly identified and characterized Sandy-related deaths; signaled excess mortality increase; informed disaster response and preparedness	Not all key data available in preliminary record (including decedent address); analytical challenges required time
Data quality	Completeness and validity of data reported in a system	Preliminary records provide complete, high-quality information on required fields; routine medical examiner reporting and text fields identified Sandy-related deaths	Limited comparability because of variation in definitions nationwide; no identification of indirectly related deaths
Timeliness	Speed between steps in a system	Existing mandates for timely reporting; no delays in medical certification	Slight delays in registration, particularly in flood-inundated areas
Stability	Reliability and availability of a system	Maintained electronic reporting; able to register deaths remotely via a Web-based system and contact users	Minor system outages
Flexibility	Adaptation to changing information needs	Existing system and staff to perform tasks; modifiable platform (e.g., preliminary records)	Delays in activation of certain features (e.g., pop-up box); other enhancements not possible on a short-term basis

^aOn the basis of the Centers for Disease Control and Prevention’s “Updated guidelines for evaluating public health surveillance systems: recommendations from the Guidelines Working Group.”¹⁴

Usefulness

NYC’s disaster mortality surveillance allowed the rapid identification and characterization of Sandy-related deaths needed to inform ongoing response. In particular, EDRS data showed that the majority of deaths were the result of drowning in coastal areas of NYC.¹⁶ Data were also used in the storm’s aftermath to estimate next-day all-cause death counts and signal an increase from prior years in deaths from all causes. Figure 2 shows the percentage difference between 2012 daily and 3-day average death counts and the average of 2010–2011 death counts from October 1 through December 31 (registered as of December 31 of each year). Preliminary findings showed that more deaths occurred in 2012 than in 2010–2011, with the percentage difference increasing up to 30% directly following Hurricane Sandy and remaining positive throughout November and December.

FIGURE 2—

Percentage differences in all-cause daily death counts in 2012 compared with the average of 2010–2011: New York City, NY, October 1–December 31, 2012.

Note. Deaths are registered as of December 31 of each year.

Detailed demographic information provided in the EDRS, such as decedent’s age and residential address, was also useful for exploring potential causes for the increase. In particular, the ability to generate census tract–level information from the decedent’s address delineated flood-affected areas in high-density areas of NYC and ultimately showed that the increase in mortality was not clustered in highly inundated areas, an important finding for ongoing response efforts. The BVS emailed reports, with a summary of key findings, to the NYC Health Department Commissioner, Office of Emergency Management, Bureau of Environmental Surveillance and Policy, and Incident Command personnel.

Several system weaknesses hindered the ability to detect trends in Sandy-related mortality. First, because of the multistep process of inputting and coding fields, some variables were not available in the preliminary record for analysis. For example, residential address is not required at the time of medical certification and is usually completed later by the funeral director; therefore, reports stratifying deaths by address (or area-based measures on the basis of geographic information system coding) were not available until after the registration of the death certificate. Similarly, whereas cause of death was available in text fields at the time of medical certification, ICD-10 codes, which are much easier to group and analyze, were not assigned until registration. Additional analytical challenges included controlling underlying differences in baseline reporting dates, accounting for population changes related to evacuation, and developing the analysis and layout for reports.

Flexibility

The NYC EDRS demonstrated flexibility in that it was adapted by the Health Department to provide disaster mortality surveillance with minimal additional time, personnel, and dedicated funds. This is primarily because the EDRS is routinely used and has established reporting procedures and relationships with data providers. The BVS could enhance and modify the system to take into account emerging storm-related conditions. For instance, to ensure the referral of injury-related fatalities to the OCME, the BVS activated a pop-up notice that displayed in the EDRS when the medical provider entered cause of death information.

In response to growing concerns about routine and storm-related delays in death registration, the BVS also modified routine Structured Query Language queries to analyze preliminary death records. Because deaths are medically certified within 24 hours, preliminary records provided almost next-day estimates of death counts, which were valuable for situational awareness. Although the EDRS allowed pop-up notices, the Sandy message was not activated until 20 days after the storm because of staff oversight during a chaotic time. Additional EDRS enhancements were not possible because of the time and money required to submit a request to the system vendor or obtain Board of Health approval, which is required for all NYC vital event registration form changes.

Stability

System stability refers to the ability to collect, manage, and provide data that are reliable and available when needed. Despite major citywide and internal Health Department disruptions, including the closure of the building where registration staff process death records, the EDRS remained stable, with only minor outages or interruptions to death reporting during and after Hurricane Sandy. The Web-based platform of the EDRS allowed the BVS to maintain operations via remote registration, even as the main building closed and staff members were relocated.

Updated BVS contact information was disseminated to facilities by e-mail using the routine EDRS notice list or by phone for selected sites in affected areas, allowing the BVS to provide technical assistance to system users. The proportion of paper certificates received the day of the storm decreased slightly to 4.4%, and increased the day after to 8.0%; however, 6.5% of deaths on average were reported on paper certificates 10 days following the storm, which is similar to the annual average.

Timeliness

Timeliness refers to the speed between system steps; in this case the time between when the death occurred and when it was medically certified and then registered in the EDRS. Compared with the same 2-week period in 2010 and 2011, we found no significant difference in the median time for medical certification in 2012 (Table 2). However, there was a statistically significant difference in the median registration time, from 2.34 and 2.27 days in 2010 and 2011, respectively, to 3.12 days in 2012, corresponding to a 35.4% increase. Although almost all boroughs and census tracts were significantly different in 2012 from 2010 and 2011, the increase in median registration time in 2012 was particularly high among deaths reported in Staten Island (54.6%) and among records in which the decedent’s address of residence was heavily flooded (87.9%).

TABLE 2—

Timeliness Between the Date of Death and Medical Certification and the Date of Death and Registration by Select Characteristics: New York City, October 29–November 10, 2010–2012

	Median Time Between Death and Medical Certification, Days			Median Time Between Death And Registration, Days
Variable	2010	2011	2012	2010	2011	2012
All	0.30	0.28	0.36	2.34	2.27	3.12a,b
Sandy-related
Yes	. . .	. . .	2.14	. . .	. . .	5.81
No	. . .	. . .	0.33	. . .	. . .	3.08
Medical examiner case
Yes	1.29	1.44	1.64a,b	3.99	4.15	4.89
No	0.20	0.19	0.22	2.16	2.03	2.92a,b
Borough of report
Manhattan	0.40	0.40	0.40	2.20	2.43	3.27a,b
Brooklyn	0.26	0.25	0.31	2.71	2.65	3.44b
Bronx	0.27	0.25	0.30	2.50	2.45	3.19a,b
Queens	0.30	0.24	0.40	2.02	1.97	2.76a,b
Staten Island	0.25	0.17	0.59b	2.00	1.83	2.96a,b
Census tract flooded, %c
0	0.31	0.28	0.33	2.52	2.42	3.09a,b
1–9	0.43	0.40	0.34	2.62	2.58	3.58
10–75	0.32	0.29	0.47	2.00	2.05	3.11a,b
> 75	0.18	0.19	0.64a	1.75	1.88	3.41a,b
Method of report
Paper	0.39	0.32	0.49	2.49	2.21	3.02
Electronic	0.30	0.27	0.35	2.30	2.28	3.14a,b

^aWilcoxon signed-rank test; statistical difference from 2010 at P < .05.

^bWilcoxon signed-rank test; statistical difference from 2011 at P < .05.

^cOn the basis of decedent’s residential address and Federal Emergency Management Agency maps.¹⁵

Sandy-related cases investigated by the OCME were registered in 5.81 days, compared with 3.08 days for all other deaths. This is approximately 1 day slower than the time required for all medical examiner cases to be registered during this time (4.89 days). The Sandy-related infrastructure disruptions across the city could have extended the time needed for death scene investigation and funeral arrangements in affected areas.

Quality

In terms of completeness, we found that the proportion of preliminary death records missing observations from November 12 through December 30, 2012, varied by item. Decedent name, sex, date of death, and cause of death, which are required when the death is medically certified, were nearly 100% complete (Table 3). Decedent residential address fields, which were not required until registration, were missing in approximately 80% of preliminary records. The decedent’s date of birth was also missing in 79% of these records. When available on both preliminary and registered records, we found on average a high percentage of agreement (> 97%). Minor spelling mistakes and first and last name reversal were most common, indicating that preliminary records provide complete and high-quality information in fields required at the time of medical certification.

TABLE 3—

Completeness and Agreement of Preliminary Records and Registered Death Records: New York City, November 12–December 30, 2012

	Completeness,a No. Present (%)
Select Death Certificate Fields	Certified	Registered	No. Agreed (%)b
Name
First	5511 (99.5)	5511 (99.5)	5420 (98.3)
Lastc	5536 (100.0)	5536 (100.0)	5447 (98.4)
Sexc	5536 (100.0)	5536 (100.0)	5522 (99.7)
Date of birth	1162 (21.0)	5534 (100.0)	1101 (94.8)
Date of deathc	5536 (100.0)	5536 (100.0)	5529 (99.9)
Type of place of deathc	5536 (100.0)	5536 (100.0)	5532 (99.9)
Manner of death	5243 (94.7)	5264 (95.1)	5235 (99.8)
How injury occurred	142 (2.6)	149 (2.7)	137 (96.5)
Cause of death
Part 1A	5536 (100.0)	5536 (100.0)	5370 (97.0)
Part 1B	3455 (62.4)	3486 (63.0)	3313 (95.9)
Part 1C	1353 (24.4)	1385 (25.0)	1294 (95.6)
Part 1D	279 (5.0)	279 (5.0)	266 (95.3)
Part 2	1735 (31.3)	1764 (31.9)	1664 (95.9)
Place of residence
Zip code	1047 (18.9)	5482 (99.0)	1027 (98.1)
City	1137 (20.5)	5534 (100.0)	1095 (96.3)
County	1133 (20.5)	5516 (99.6)	1113 (98.2)
State	1136 (20.5)	5523 (99.8)	1130 (99.5)

Note. The sample size was n = 5536.

^aCompleteness refers to the absence of missing or unknown values for a select field.

^bThe frequency of records that agreed (no. agreed) divided by the number of certified records present (no. present) by item.

^cFields required at the time a record is medically certified.

We also found that conducting text string searches and scanning media reports directly following the storm verified, but did not identify, additional Sandy-related deaths, supporting the validity of routine processes of medical examiner referral. Using specific injury terms, the text string search identified 86 deaths certified as of November 12, 2012. Approximately 46 cases had been previously identified or investigated by the OCME. Of the remaining 40 death records, only 11 deaths had “accident” as the manner of death, and after review none appeared to be directly related to Sandy. The BVS referred 3 additional deaths identified by the media as related to Hurricane Sandy; however, the OCME determined that these deaths did not meet their Sandy-related definition.

Of note, all 44 deaths identified by the OCME had “Hurricane Sandy” in the “Describe How This Injury Occurred” field, facilitating identification and coding of disaster-related deaths even without direct OCME notification. Therefore, the ability to identify disaster-related deaths after the storm using cause of death and injury death certificate text fields proved to be a system strength. However, additional, indirectly related deaths, which are potentially important for public health and disaster response, were not identified through this process and would require additional resources, such as manual case review. Furthermore, the definition of a disaster-related death may differ by jurisdiction, limiting comparability across affected areas.

DISCUSSION

The NYC Health Department’s EDRS successfully provided disaster-related mortality surveillance following Hurricane Sandy. Despite unprecedented disruptions in city infrastructure, the BVS maintained system stability by establishing off-site registration and communicating directly with facilities and data providers. As a result, minimal delays in medical certification occurred compared with prior years. Registration during this time was delayed compared with prior years, particularly in heavily flooded areas; however, the BVS counteracted this by extracting preliminary records from the EDRS, which because of reporting mandates are available within 24 hours of death. Additional EDRS enhancements in response to changing conditions and needs were also possible with minimal added resources.

Overall, the mortality surveillance system rapidly identified and characterized Sandy-related deaths, signaled an increase in excess mortality, and informed disaster response and preparedness. The system also had some weaknesses, in particular the lack of preestablished methods and reports for disaster-related mortality surveillance, which delayed surveillance of all-cause mortality. Although the system was flexible to some changes, others were not possible because of the time and money needed to add items to the electronic system. Finally, the system was limited in its ability to identify indirectly related deaths in a timely way that would be comparable with that of other jurisdictions.

Nationwide, real-time EDRS disaster mortality surveillance is limited. EDRSs are still relatively new and have yet to be implemented in 11 jurisdictions. Moreover, complete electronic reporting is low; as of April 2014, only 15 jurisdictions reported 80% or more of their death records entirely electronically.¹⁷ Vital statistics departments are also lacking many of the resources needed to conduct this type of surveillance, including local cause of death coding, real-time geocoding, and data extraction.

Historically, the Centers for Disease Control and Prevention has partnered with the American Red Cross to address these gaps in disaster-related mortality nationwide, and in the past decade this system has provided mortality information for several disasters.^18–21 The Centers for Disease Control and Prevention also designed a disaster-related mortality surveillance form, which can be implemented by health departments after a disaster to collect basic mortality information.²² Following Hurricane Hugo in Texas in 1989, this form was more successful in identifying hurricane-related deaths than was the paper-based death registration system; however, the form required special training for providers and resources for mailing or faxing and was not significantly faster than was routine death registration (16 vs 14 days, respectively).¹² As EDRSs expand nationally, there may be opportunities to enhance these existing systems.

On the basis of this evaluation, we developed several recommendations to improve disaster-related mortality surveillance. First, further assessment of indirectly related deaths and excess mortality is needed to determine gaps in current surveillance activities and explore potential reasons for the increase in mortality. Second, standardized statistical methods for estimating expected and excess mortality and reporting to emergency response personnel should be developed that are customized for various emergency response scenarios and account for the registration lag.²³ This will improve the timelines and usability of mortality surveillance reports and allow greater flexibility in other disasters. Third, advanced text string search methods, similar to syndromic surveillance, could improve case identification and categorization of causes before registration and ICD-10 assignment. These methods could also be used to identify indirectly related deaths, which would narrow or eliminate the need to manually review records postdisaster.

Fourth, a continued focus on the quality and timeliness of cause of death reporting is essential for identifying disaster-related deaths during a disaster. To standardize definitions and practices across jurisdictions, national guidelines and best practices for disaster-related death investigation are also needed. As specifications for newer EDRSs are developed, further adaptations should be explored. For example, decedent address information could be added earlier in the certification process to facilitate earlier geographic analysis, a key factor in most disasters. Drop-down menus for reporting specific health events, such as those used in Ohio during the 2009–2010 influenza A H1N1 pandemic, could further improve timeliness and case ascertainment.²⁴ Finally, periodic assessment of the mortality surveillance system, particularly after disasters, should be conducted. This would allow the BVS to evaluate the system’s performance and monitor the success of any improvements.

Limitations

There are 3 limitations to this evaluation that should be considered. First, we did not evaluate the system’s sensitivity in detecting Sandy-related deaths, mainly because no external data source exists for comparison in this population. Second, we did not interview data providers, including physicians and funeral directors, although they are a critical part of the registration process. Finally, we were unable to accurately estimate the direct financial costs of using the system for surveillance.

Conclusions

Vital statistics are considered a core epidemiological tool for understanding trends in population health, developing and evaluating public health interventions, and responding to emerging threats. Over the past century, vital registration has become routine in every jurisdiction nationwide; however, more work is needed to implement, develop, and use electronic systems to their full capacity. EDRSs provide a promising opportunity to conduct timely surveillance that was not possible with paper records and have the potential to support a wide range of public health concerns and responses, including influenza- and drug overdose–related surveillance and mass fatality events. This evaluation provides just 1 example of using an EDRS for disaster-related mortality surveillance.

We found that NYC’s EDRS allowed the successful surveillance of mortality during and after Hurricane Sandy, providing useful, timely information for response efforts. Our system evaluation also documented areas of weakness that should be addressed to improve surveillance capabilities for future public health emergencies. These findings can support local vital jurisdictions as they develop EDRSs, work toward complete electronic reporting, and build surveillance capabilities. Public health officials at the local, state, and federal level should partner with vital records and statistics offices to support expansion of EDRSs, develop mortality surveillance methods and uses, and promote ongoing evaluation.